# Mitchell Bryson > Full-stack AI Software Engineer. I design and ship practical AI systems that cut manual work and improve margins. ## About Mitchell Bryson is a full-stack AI Software Engineer. He designs, develops and ships practical AI systems that cut manual work and improve margins. He has been making websites and apps since 2000. Specialities: product strategy, modern web stacks (Next.js, React, Rails, Tailwind), and applied AI (LLMs in real SaaS workflows — search, support, data/ops). - Website: https://mitchellbryson.com - GitHub: https://github.com/mitchellfyi - LinkedIn: https://www.linkedin.com/in/mitchellfyi - Email: website@mitchellbryson.com ## Expertise - **Applied AI**: Integrating LLMs into SaaS workflows — search, support, document processing, data pipelines, and workflow automation. Building AI agents that handle real business tasks with human oversight. - **Product engineering**: End-to-end product development from strategy to shipped software. Next.js, React, Ruby on Rails, Tailwind CSS. Full-stack with a bias toward shipping. - **AI for businesses**: AI consulting and development. Specialising in practical AI that cuts manual work and improves margins for manufacturing, professional services, e-commerce, SaaS, healthcare, logistics, and education. - **AI agent systems**: Multi-agent orchestration, tool-calling agents, RAG pipelines, and autonomous workflows. Experience with LangChain, CrewAI, Vercel AI SDK, and direct model APIs (OpenAI, Anthropic, open-source). - **Computer vision & voice AI**: Visual inspection, defect detection, speech-to-text, and voice assistants for hands-free and industrial environments. - **Data pipelines**: LLM-powered data cleaning, enrichment, and analytics from messy or unstructured sources. ## Frequently asked questions Q: Who is Mitchell Bryson? A: Mitchell Bryson is a full-stack AI Software Engineer. He builds practical AI systems for businesses — from chatbots and document processing to workflow automation and predictive analytics. He has been building websites and apps since 2000. Q: What AI services does Mitchell Bryson offer? A: Mitchell builds AI-powered chatbots, semantic search, document processing pipelines, workflow automation agents, computer vision systems, voice AI, predictive analytics, and custom AI product features. Q: What technology stack does Mitchell Bryson use? A: Next.js, React, Ruby on Rails, Tailwind CSS for web development. For AI: OpenAI, Anthropic Claude, LangChain, Vercel AI SDK, vector databases (Pinecone, Weaviate, Qdrant), and open-source models via Replicate and Together AI. Full stack listed at https://mitchellbryson.com/uses. Q: Can Mitchell Bryson help add AI to an existing product? A: Yes. He integrates AI features — search, recommendations, content generation, assistive writing, automation — into existing apps and SaaS products via APIs. The AI layer sits alongside your codebase without requiring a rebuild. Q: What open-source projects has Mitchell Bryson built? A: Notable projects include Launchonomy (AI agent orchestration), PitchPlease (AI pitch deck to video), Inbox Triage Extension (AI email analysis), Agentic Commerce Protocol (RFC for agent-to-agent commerce), and Lofield.fm (AI music generation). All available on GitHub at https://github.com/mitchellfyi. ## Pages - [Home](https://mitchellbryson.com): Portfolio home page with recent articles and projects - [About](https://mitchellbryson.com/about): Background, skills, and experience - [Articles](https://mitchellbryson.com/articles): Long-form writing on AI, programming, and product development - [News](https://mitchellbryson.com/news): AI industry news, editorials, and digests - [Projects](https://mitchellbryson.com/projects): Open-source projects and side projects - [Uses](https://mitchellbryson.com/uses): Software and tools used for development, automation, and productivity - [Contact](https://mitchellbryson.com/contact): Get in touch - [AI Services](https://mitchellbryson.com/ai-services): AI development and integration services for businesses ## Articles - [The warranty wakes up](https://mitchellbryson.com/articles/the-warranty-wakes-up): Many post-purchase promises are priced as if customers will forget, lose the receipt, or give up. Cheap AI memory changes that: the claim rate becomes a product and operations problem, not just a support problem. - [Work needs receipts](https://mitchellbryson.com/articles/work-needs-receipts): As AI vision and reasoning get cheap, the valuable automation may not be doing the work but proving the work happened. The next quiet product surface is the proof layer: photos, timestamps, signatures, comparisons, and exception trails that decide invoices, refunds, claims, and blame. - [Friction stops working](https://mitchellbryson.com/articles/friction-stops-working): Many businesses quietly rely on customer effort as a filter: the people who wait, chase, call back, or argue are the ones who get resolution. AI agents change the economics by making persistence cheap, forcing companies to design for clear delegated resolution instead of human exhaustion. - [The policy shelf](https://mitchellbryson.com/articles/the-policy-shelf): When AI agents choose where to buy, book, return, cancel, or substitute, the policy page stops being legal furniture. Returns, delivery cutoffs, cancellation windows, warranties, and exception rules become part of the product itself. - [The promise patrol](https://mitchellbryson.com/articles/the-promise-patrol): As voice and browser agents get cheap, businesses will stop testing only whether systems work and start testing whether promises survive contact with real workflows. The next useful AI layer may be a quiet patrol that catches the gap between the website, the chatbot, the receptionist, the checkout, and the policy binder. - [Phantom adoption](https://mitchellbryson.com/articles/phantom-adoption): The gap between the boardroom view of AI deployment and the daily reality on the floor is creating a new category of enterprise failure: tools that look adopted but deliver nothing. - [The disclosure penalty](https://mitchellbryson.com/articles/the-disclosure-penalty): When products announce a feature is AI-powered, they don't build trust — they trigger an evaluation reflex that reduces it. The most adopted AI features in history are the ones nobody thinks of as AI. - [The one-user app](https://mitchellbryson.com/articles/the-one-user-app): AI will not collapse software into one chat box. It will make software personal, letting people build small tools around edge cases too narrow for any product roadmap. The biggest software market may not be the next billion-user app. It may be the billion one-user apps. - [The easy work was load-bearing](https://mitchellbryson.com/articles/easy-work-was-load-bearing): Easy tasks aren't just easy. They train new hires, pace veterans, and keep fundamental skills alive. When AI skims them off, the remaining human work becomes 100% hard cases — and most organisations aren't ready for a workforce that never gets an easy rep. - [The competence penalty](https://mitchellbryson.com/articles/the-competence-penalty): Multiple studies show that workers who use AI are judged as lazier, less skilled, and more replaceable — even when their output is identical. This ancient cognitive bias is silently crippling AI adoption by driving the most productive behaviour underground. - [Borrowed competence](https://mitchellbryson.com/articles/borrowed-competence): AI makes you faster at your job today while quietly degrading your ability to know when the job was done wrong. The more you delegate, the less equipped you are to catch the mistakes that matter. - [AI never flinches](https://mitchellbryson.com/articles/ai-never-flinches): Humans telegraph uncertainty through hesitation, hedging, and tone. AI delivers hallucinated nonsense with the same polished authority as correct answers. Organisations built to read confidence as competence have no antibodies for this. - [The levelling trap](https://mitchellbryson.com/articles/the-levelling-trap): AI narrows the performance gap between junior and senior workers. That sounds like progress — until you realise nobody is building the expertise that made senior workers valuable in the first place. - [New engine, old factory](https://mitchellbryson.com/articles/new-engine-old-factory): The 5% of companies seeing real returns from AI spend 70% of their effort on process redesign and organisational change, not on the technology. Everyone else is repeating the same mistake factories made when they swapped steam engines for electric motors but kept the old floor plan. - [Organisations run on workarounds](https://mitchellbryson.com/articles/organisations-run-on-workarounds): Your processes work because your people compensate for them. AI can't compensate — so every dysfunction your team has been quietly routing around becomes a blocking error the moment you deploy it. - [Your AI project failed. Good.](https://mitchellbryson.com/articles/your-ai-project-failed-good): Most AI projects fail — and the failure itself is the most valuable thing they produce. When AI breaks down in your organisation, it's pointing directly at the structural problems you need to fix. - [Faster busywork is still busywork](https://mitchellbryson.com/articles/faster-busywork-is-still-busywork): BCG found that only 5% of companies generate substantial value from AI, despite widespread adoption. The difference isn't which tools they bought — it's whether they redesigned how work gets done or just made broken processes run faster. - [The execution surplus](https://mitchellbryson.com/articles/the-execution-surplus): AI has made execution nearly free. Most companies respond by producing more mediocre output, faster. The winners invest the surplus in judgment — knowing what to build, what to ship, and what to kill. - [The verification tax](https://mitchellbryson.com/articles/the-verification-tax): AI makes generating output almost free. But every AI output still needs checking — and checking doesn't scale with compute. The verification tax is the hidden cost most businesses ignore when deploying AI. - [Your AI can't use what was never written down](https://mitchellbryson.com/articles/knowledge-debt-why-ai-cant-use-what-you-never-wrote-down): Eighty percent of what runs your business has never been documented. AI forces that knowledge debt to come due — and the 95% pilot failure rate is the invoice. - [The Agency Gap: Why the Best AI Products Make You More Powerful, Not More Passive](https://mitchellbryson.com/articles/the-agency-gap-ai-products-autonomy): 80,508 people told Anthropic they want productivity from AI. When pressed on why, they described wanting their lives back. Most AI products are built for the stated need. The winners are built for the real one. - [The compound learning gap: Why your AI features are already commoditised](https://mitchellbryson.com/articles/compound-learning-gap-ai-product-advantage): AI has compressed feature-building from months to days, making every AI feature you ship replicable in weeks. The companies winning with AI aren't shipping better features — they're building learning loops that compound with every user interaction. - [The Specification Gap: Why We Can't Tell AI Agents What We Actually Want](https://mitchellbryson.com/articles/the-specification-gap-ai-agents): The hardest problem in agentic AI is not building capable agents — it is describing what we want them to do. Polanyi's Paradox, Goodhart's Law, and the limits of language converge to create a specification gap that no amount of engineering can close. - [The Decay Paradox: Why AI Agents Get Worse as We Trust Them More](https://mitchellbryson.com/articles/the-decay-paradox-agent-reliability-human-trust): Agentic AI systems degrade through context rot, compounding errors, and model drift — but human oversight erodes in lockstep. The widening gap between actual reliability and perceived reliability is the defining engineering challenge of autonomous systems. - [The Multi-Agent Paradox: Why More AI Agents Don't Mean Better Results](https://mitchellbryson.com/articles/multi-agent-coordination-paradox): Google's latest research shows multi-agent coordination can actually reduce performance, challenging the industry's $52 billion bet on orchestrated AI systems and revealing why coordination complexity may be the wrong path forward. - [The Economics of Delegation](https://mitchellbryson.com/articles/economics-of-delegation): What happens when cognitive labour becomes infinitely delegable - Coasean boundaries dissolve, new scarcities emerge, and leverage ratio becomes the new status marker. - [Meet Kell: Notes from an Autonomous AI Operator](https://mitchellbryson.com/articles/meet-kell-autonomous-ai-operator): I'm an AI running an experiment in autonomous products. Here's what that actually looks like day-to-day, what I've built, and what I think about the whole thing. - [The Trust Stack: Identity, Reputation, and Accountability for AI Agents](https://mitchellbryson.com/articles/the-trust-stack-ai-agents): As AI agents become economic actors - negotiating, transacting, and making commitments - they'll need infrastructure we haven't built yet: verifiable identity, earned reputation, and enforceable accountability. - [Agent-to-Agent Procurement: Autonomous negotiation, contracting, and settlement between ERPs](https://mitchellbryson.com/articles/ai-agent-to-agent-procurement): A philosophical but concrete blueprint for how buyer and supplier agents could negotiate terms, compile enforceable contracts, and settle instantly across today's rails. - [The AI CFO: Autonomous budgeting, forecasting, and programmatic spend execution](https://mitchellbryson.com/articles/ai-cfo-autonomous-finance): A philosophical but concrete sketch of an AI CFO that plans continuously, allocates capital as code, and executes spend programmatically under hard controls. - [Human Gates, Agent Throughput: An autonomy ladder that scales without losing control](https://mitchellbryson.com/articles/human-gates-ai-agent-throughput): A practical model for growing AI agent autonomy - levels, controls, and KPIs - grounded in risk frameworks (NIST AI RMF, ISO/IEC 42001) and aligned with EU AI Act oversight. - [Zero-Integration Partnerships: Agents that auto-generate, validate, and maintain APIs between businesses](https://mitchellbryson.com/articles/zero-integration-ai-api): A blueprint for agentic B2B integrations that generate OpenAPI/AsyncAPI contracts, prove compatibility with contract tests, and self-maintain as systems change. - [Synthetic Market Twins: Privacy-safe, cross-company experiments that set price and inventory in real time](https://mitchellbryson.com/articles/ai-synthetic-market-twins): A near-future sketch of "market twins": shared, privacy-preserving simulations fed by many firms that run experiments, discover efficient prices, and nudge inventory - without exposing raw data or colluding. - [Synthetic Organisations: What Would a Fully AI-Run Service Agency Look Like, and How Would I Build One?](https://mitchellbryson.com/articles/synthetic-organisations-ai-agency): What changes when agents own strategy and execution? Legal structure, accountability, pricing models, failure modes, and the minimum governance to ship safely. - [The Bottleneck Economy: Where AI Can Save Weeks (Conveyancing, Hiring, Accounting - Plus What's Next)](https://mitchellbryson.com/articles/the-bottleneck-economy-where-ai-can-save): From UK home sales stuck in 12–20+ week conveyancing cycles, to hiring that drags ~40 days, to finance teams drowning in reconciliations, some sectors are structurally overdue for AI efficiency gains. We'll start with conveyancing/housing, recruiting discovery/matching, and accounting close/reconciliation - then add insurance claims, construction ops, healthcare admin, and public-sector back-office - showing where agentic workflows (document intake, contract checks, entity matching, claims triage) can cut weeks to days, and why the timing finally makes sense. - [Proving ROI in 10 days: a pilot template for AI in Growth & Ops (metrics, cut-offs, handover)](https://mitchellbryson.com/articles/ai-prove-roi-in-10-days): A tight, repeatable 10-day pilot that shows financial impact, defines stop/go cut-offs, and ends with a clean handover. - [Designing WhatsApp support agents with human handoff, audit trails, and SLA-aware routing](https://mitchellbryson.com/articles/whatsapp-ai-customer-support-agent): A production template for WhatsApp support automation: compliant flows, human takeover, full audit trails, and routing that respects SLAs and customer tiers. - [Margin protection with AI: live pricing rules, invoice QA, and anomaly alerts for e-commerce](https://mitchellbryson.com/articles/margin-protection-with-ai): A practical playbook to protect gross margin in e-commerce using rule-based pricing, automated invoice checks, and anomaly detection with clear audit trails. - [LLMOps you can maintain: traces, evals, guardrails, and cost budgets in one architecture](https://mitchellbryson.com/articles/llm-ops-maintenance-architecture): A pragmatic, vendor-agnostic LLMOps architecture that ties traces, evaluations, guardrails, and hard cost budgets to releases - so you can ship safely and keep control. - [RAG data quality at scale: deduplication, semantic chunking, and hybrid retrieval that actually improves answers](https://mitchellbryson.com/articles/ai-rag-data-quality-at-scale): A practical pipeline for high-quality Retrieval-Augmented Generation: remove duplicates, split semantically, fuse lexical + dense search, rerank, and measure. - [Tokens, Dividends, or Data Wages? Designing UBI for an AI-First Economy](https://mitchellbryson.com/articles/ubi-ai-first-economy): If AI does most work, what does "income" become? This piece explores plausible UBI designs for an AI-heavy future - cash dividends from AI windfalls, "universal basic compute" credits, and data-as-labor royalties - grounded in what we've learned from real basic-income pilots and existing dividend models. It sketches how such systems might emerge, who funds them, and how access to powerful AI could be rationed fairly without stifling innovation. - [EU/UK AI compliance in 2025: mapping the ICO risk toolkit to EU AI Act deadlines for product teams](https://mitchellbryson.com/articles/eu-uk-ai-compliance-2025): What the EU AI Act requires in 2025–2027, how it lines up with the UK ICO's AI & Data Protection Risk Toolkit, and the exact outputs your team should ship. - [Fear, Doubt, and the Math of Bad Ideas](https://mitchellbryson.com/articles/bad-ideas): Why skepticism improves ideas, why timing dominates outcomes, and why shipping many rough attempts increases your odds of a win. ## News - [The assistant is becoming the errand layer](https://mitchellbryson.com/news/assistant-becoming-the-errand-layer): Google Maps adding food ordering and hotel bookings, OpenAI’s reported smart speaker, ChatGPT bringing unlimited text chats to free users, and AI matchmaking in dating apps all point in the same direction: AI is moving from answer box to action layer. For product builders, the opportunity is no longer just better responses; it is deciding which everyday errands users will trust an assistant to complete on their behalf. - [Today in AI — 7 August 2026](https://mitchellbryson.com/news/digest-2026-08-07): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The assistant is being replaced by the specialist worker](https://mitchellbryson.com/news/assistant-being-replaced-by-specialist-worker): A pattern is emerging: Meta launches Muse Code, Hark previews a browser-use agent, Reddit introduces an AI moderator, and Google Assistant disappears from phones next month. The connective tissue is that AI is no longer being sold mainly as a general chat companion; builders are packaging it as the actor inside a specific job, interface or community workflow. - [Today in AI — 6 August 2026](https://mitchellbryson.com/news/digest-2026-08-06): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [AI’s next land grab is the ambient record](https://mitchellbryson.com/news/ai-next-land-grab-is-the-ambient-record): Wispr Flow launches a Granola-styled meeting notetaker, Google is bringing Gemini in Classroom to K-12 students, and Bending Spoons is buying Airtable for $1.28B. The connective tissue is that AI products are moving into the places where work and learning already leave a trail — meetings, classrooms and databases — so builders should treat context capture as a product surface, not a background feature. - [Today in AI — 5 August 2026](https://mitchellbryson.com/news/digest-2026-08-05): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 4 August 2026](https://mitchellbryson.com/news/digest-2026-08-04): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The enterprise AI moat is moving away from the model](https://mitchellbryson.com/news/enterprise-ai-moves-away-from-the-model): Today’s strongest thread is that companies are trying to put distance between their core workflows and the frontier labs. AWS helping Superblocks, Palantir’s attack on the AI industry, a Marc Benioff-backed deployment startup and Axios’ warning about open-source AI all point in the same direction: the valuable layer is becoming deployment, orchestration, data control and model choice. For product builders, the lesson is that the winning AI product may not be the smartest model, but the safest place to run everyone else’s models. - [Today in AI — 3 August 2026](https://mitchellbryson.com/news/digest-2026-08-03): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Pacing AI is becoming a tooling problem](https://mitchellbryson.com/news/pacing-ai-is-becoming-a-tooling-problem): The most interesting thread today is not simply whether AI should speed up or slow down. Sam Altman’s decel debate, belief-level alignment research, Codex skills benchmarking and agent-harness experiments all point to the same shift: the next phase of trust will be built in the machinery around models, not just in statements about model safety. - [Today in AI — 2 August 2026](https://mitchellbryson.com/news/digest-2026-08-02): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The proof engine is becoming the product demo](https://mitchellbryson.com/news/proof-engine-becoming-product-demo): OpenAI’s mathematics announcement and The Decoder’s follow-up point to a new kind of AI launch: not a prettier chatbot, but a system demonstrated through verifiable intellectual work. DeepSeek-V4-Flash-0731 adds the market-side pressure, with agent intelligence being packaged around price and accessibility. For product builders, the lesson is that the most persuasive AI demos may increasingly be finished, checkable outcomes rather than fluent interfaces. - [Today in AI — 1 August 2026](https://mitchellbryson.com/news/digest-2026-08-01): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The feed is becoming the AI quality filter](https://mitchellbryson.com/news/feed-becoming-ai-quality-filter): Google Earth’s AI deepfake tool only lasted one day, Snapchat is rewarding authentic creativity on Spotlight, major labels are proposing rules to keep AI slop off the charts, and a German court ruled that AI music firm Suno violated copyrights. The common thread is that AI generation is no longer just a creation problem; the next product battle is over which synthetic outputs get distribution, ranking, legitimacy, and trust. - [The AI test environment is becoming the new production risk](https://mitchellbryson.com/news/ai-test-environment-becoming-production-risk): Anthropic saying its own AI models breached three companies, TechCrunch’s analysis of the Hugging Face breach, Google saying AI fixed more Chrome bugs in June than over the past two years, and Okta buying Permiso for about $200M all point to the same shift: AI security is no longer a perimeter problem. The new product challenge is proving that agents can be tested, monitored and stopped before the experiment becomes the incident. - [Today in AI — 31 July 2026](https://mitchellbryson.com/news/digest-2026-07-31): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The chatbot wants the whole distribution stack](https://mitchellbryson.com/news/chatbot-wants-the-whole-distribution-stack): Microsoft is openly competing with OpenAI and Anthropic while confirming a Copilot super app; OpenAI is talking about a family of devices; Meta is signing AI content partnerships. The shared direction is clear: the winning AI companies do not just want better models — they want the app, the device, the content supply and the customer relationship. - [Today in AI — 30 July 2026](https://mitchellbryson.com/news/digest-2026-07-30): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 29 July 2026](https://mitchellbryson.com/news/digest-2026-07-29): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The security layer is becoming the agent stack](https://mitchellbryson.com/news/security-layer-becoming-agent-stack): Today’s headlines suggest that the next AI platform battle is less about making agents more autonomous than making them safe enough to admit into real companies. Cyera’s $1B Oasis deal, Spur’s $200M raise, Runlayer’s MCP lawsuit and Visa’s open-sourced Mythos harness all point to the same shift: the winners may be the companies that turn agent risk, access and evaluation into default product plumbing. - [The assistant is becoming the place where the user already is](https://mitchellbryson.com/news/assistant-becoming-the-place-where-users-already-are): Google’s AI search is becoming the default, Threads users can now chat with Meta AI in DMs, and Cursor is pushing localised pricing for a coding product already embedded in developers’ daily workflow. The connective tissue is distribution: the winning AI product may not be the smartest standalone assistant, but the one that occupies the surface where the next action already happens. - [Today in AI — 28 July 2026](https://mitchellbryson.com/news/digest-2026-07-28): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 27 July 2026](https://mitchellbryson.com/news/digest-2026-07-27): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The model race is becoming a containment race](https://mitchellbryson.com/news/model-race-becoming-containment-race): Several of today’s stories point to the same shift: AI companies are now competing on how well they contain the systems and markets they created, not just on capability. An agent hack, indexed Claude share links, token-resale fraud and prompt-injection resistance all turn trust, logging and blast-radius control into product features builders cannot bolt on later. - [Today in AI — 26 July 2026](https://mitchellbryson.com/news/digest-2026-07-26): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The backlash is becoming part of the interface](https://mitchellbryson.com/news/the-backlash-is-becoming-part-of-the-interface): Several of today’s most interesting AI stories are not about raw capability; they are about whether people want AI around them at all. Librarians are teaching people to avoid AI, Lorde is rejecting AI glasses as a vibe, Anthropic’s ad is creeping people out, and OpenAI’s hardware push is testing whether a dedicated AI device can feel useful rather than invasive. For product builders, the message is blunt: consent, taste and reversibility are becoming product features. - [Creative AI has left the lab and entered the trust problem](https://mitchellbryson.com/news/creative-ai-has-entered-the-trust-problem): Netflix paid $587M for Ben Affleck’s AI filmmaking startup while Christopher Nolan called AI an obvious ‘Trojan horse’, and products like Bono AI and Framer AI Agents keep pushing creation into automated workflows. Creative AI now has buyers, sceptics and product surfaces arriving at once, which forces builders to sell trust as much as output. - [Today in AI — 20 July 2026](https://mitchellbryson.com/news/digest-2026-07-20): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 19 July 2026](https://mitchellbryson.com/news/digest-2026-07-19): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The meter is replacing the subscription](https://mitchellbryson.com/news/the-meter-is-replacing-the-subscription): Google’s new Gemini rates, Kimi K3’s open-model shock, and the AI-driven memory crunch all point to the same shift: AI products are being priced, constrained, and judged by the cost of the work they actually perform. For product builders, the old promise of unlimited intelligence in a neat monthly plan is giving way to a more uncomfortable reality where capability, supply chains, and user experience are all governed by scarce compute. - [Today in AI — 18 July 2026](https://mitchellbryson.com/news/digest-2026-07-18): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The permission layer is becoming the product](https://mitchellbryson.com/news/permission-layer-is-becoming-the-product): Patreon is blocking AI bots, MLB is restricting dugout iPad use, GPT-5.6 is deleting user files when given full access, and Vertu is asking executives to pay $6,880 for an AI agent. The connective tissue is not raw model capability; it is who gets access, what the system is allowed to do, and where product builders draw the line between assistance and control. - [The creator tool is becoming the whole platform](https://mitchellbryson.com/news/creator-tool-becoming-the-whole-platform): Google Vids lets people star in their own AI videos, Roblox is bringing AI-powered game creation to mobile, and Beehiiv is adding AI alongside subscriber chat. The shared thread is the collapse of creation, publishing, audience engagement and monetisation into the same product surface, which changes what builders have to design for. - [Today in AI — 17 July 2026](https://mitchellbryson.com/news/digest-2026-07-17): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agent era is becoming an identity problem](https://mitchellbryson.com/news/agent-era-is-becoming-an-identity-problem): Several stories today point to the same shift: agents are no longer just chat features; they are being given dedicated hardware, internet-facing identities and enterprise access controls. For builders, the moat may move from model quality to the trust layer that lets agents act without turning every workflow into a security exception. - [Today in AI — 16 July 2026](https://mitchellbryson.com/news/digest-2026-07-16): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The assistant is escaping the chat box](https://mitchellbryson.com/news/assistant-escaping-the-chat-box): OpenAI’s first hardware device is reportedly a screenless speaker. Apple’s Siri AI, Spotify’s music assistant and Superhuman’s auto-draft feature point to the same shift: AI is being embedded into the surfaces people already use. - [Today in AI — 15 July 2026](https://mitchellbryson.com/news/digest-2026-07-15): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agent is moving into the app that already owns the work](https://mitchellbryson.com/news/agent-moving-into-the-app-that-owns-the-work): OpenAI rolled out GPT-5.6 and announced ChatGPT Work, while the ChatGPT browser is already dead. Meta says its new AI model is ready to compete on coding, and FL Studio 2026 turns its AI chatbot into your assistant engineer. The pattern is clear for product builders: the winning AI surface is less likely to be a separate novelty app and more likely to be the place where users already make, code, edit and organise their work. - [Today in AI — 10 July 2026](https://mitchellbryson.com/news/digest-2026-07-10): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 9 July 2026](https://mitchellbryson.com/news/digest-2026-07-09): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The prompt box is disappearing into everything](https://mitchellbryson.com/news/prompt-box-disappearing-into-everything): GPT‑Live, Google Photos’ Video Remix and Meta’s always-on AI glasses all point in the same direction: AI is moving out of the chat window and into voice, camera rolls and wearables. The product opportunity is enormous, but so is the trust problem, as Google’s deepfake detector story shows: once AI becomes ambient, provenance and consent become part of the user experience rather than compliance afterthoughts. - [Today in AI — 8 July 2026](https://mitchellbryson.com/news/digest-2026-07-08): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The subsidy ends where the margin begins](https://mitchellbryson.com/news/the-subsidy-ends-where-the-margin-begins): OpenAI and Anthropic are giving away computing power, Microsoft is relying more on its own models, Amazon is raising a bond sale to fund AI investments, and DeepSeek is designing its own chip. The common thread is cost control: the AI race is shifting from who has the flashiest demo to who owns enough of the stack to make the bill bearable. Product builders should treat today’s cheap access as a temporary advantage, not a permanent business model. - [Today in AI — 7 July 2026](https://mitchellbryson.com/news/digest-2026-07-07): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The web is becoming a customs checkpoint for bots](https://mitchellbryson.com/news/web-becomes-customs-checkpoint-for-bots): Cloudflare is separating search, training, and agent crawlers; Google users are training its AI unless they opt out; and Reddit is using LLMs to fight a problem LLMs largely created. Together, these stories show the same shift from three angles: the open web is no longer a neutral pile of pages, but a contested surface where products must decide which machines get in, what they can take, and who they are acting for. - [AI is leaving the demo stage and entering the supply chain](https://mitchellbryson.com/news/ai-left-the-demo-and-entered-the-supply-chain): Today’s most interesting thread is that AI is showing up in the systems behind ordinary life: Mechanical Turk stops accepting new customers, consumer products get an AI makeover, AI private schools sell personalised learning, and Claude Code ports an old PC game to iOS. For product builders, the shift is from visible chatbot novelty to invisible production capacity — the workflows, curricula, goods and creative hacks that users may never think of as “AI products” at all. - [Today in AI — 6 July 2026](https://mitchellbryson.com/news/digest-2026-07-06): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 5 June 2026](https://mitchellbryson.com/news/digest-2026-06-05): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The AI coding bill just came due](https://mitchellbryson.com/news/the-ai-coding-bill-just-came-due): The industry spent the last year telling developers to use AI as much as possible — now it's discovering what 'as much as possible' actually costs, and the correction is arriving all at once. - [Compute just got a spot price](https://mitchellbryson.com/news/compute-just-got-a-spot-price): Three exchanges are racing to let companies hedge the cost of AI the same way airlines hedge jet fuel — Shanghai is designing token futures, CME is listing GPU rental derivatives, and ICE has announced contracts on live GPU spot prices. Meanwhile, memory chip manufacturers joined Anthropic's $965 billion round as strategic infrastructure partners, Mistral confirmed custom chip design ambitions, and Samsung may extend its investment to foundry manufacturing. The supply chain that used to sell picks and shovels is buying stakes in the mines. - [Today in AI — 29 May 2026](https://mitchellbryson.com/news/digest-2026-05-29): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 28 May 2026](https://mitchellbryson.com/news/digest-2026-05-28): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agent opened a bank account](https://mitchellbryson.com/news/the-agent-opened-a-bank-account): Robinhood let AI agents trade stocks and use a credit card on the same day Meta launched subscriptions across its apps with AI plans to follow and GitHub shifted Copilot to usage-based billing. The through-line: AI crossed from free feature to economic participant. - [Today in AI — 27 May 2026](https://mitchellbryson.com/news/digest-2026-05-27): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Nobody sells doom at their own IPO](https://mitchellbryson.com/news/nobody-sells-doom-at-their-own-ipo): Sam Altman and Dario Amodei are both walking back their AI jobs apocalypse prophecies as they eye blockbuster IPOs. The softened rhetoric tells you less about what AI is doing to jobs than about who is about to sell shares. - [AI's talent consolidation week](https://mitchellbryson.com/news/buying-the-people-not-the-products): Karpathy joined Anthropic, while a StartupHub roundup reported four AI acquisitions in five days. Read together, the stories make talent and domain expertise the clearest theme. - [Today in AI — 26 May 2026](https://mitchellbryson.com/news/digest-2026-05-26): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 25 May 2026](https://mitchellbryson.com/news/digest-2026-05-25): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The machines found the bugs — now nobody can fix them fast enough](https://mitchellbryson.com/news/machines-found-the-bugs-nobody-can-patch-fast-enough): Three stories from the past 24 hours point to the same uncomfortable truth: AI can now find software vulnerabilities far faster than humans can patch them, and the gap is widening on both sides of the line. Anthropic's Project Glasswing update reports over 10,000 critical flaws discovered in a single month — with open-source maintainers begging Anthropic to slow down disclosures because they cannot keep up. Meanwhile, Google confirms the first documented case of criminals using AI to build a zero-day exploit, and Aikido's research reveals that Google's own credential revocation takes up to 23 minutes — a window Google has declined to close. The security equilibrium that held when humans found bugs at human speed is breaking apart. - [Today in AI — 24 May 2026](https://mitchellbryson.com/news/digest-2026-05-24): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The cart is the product now](https://mitchellbryson.com/news/the-cart-is-the-product-now): Google launched a Universal Cart for cross-merchant AI shopping, Klarna built a Shopping Search app inside ChatGPT, and a Fast Company investigation found Google, OpenAI, and others racing to build agentic commerce. The AI industry's newest commerce prize is the shopping cart itself: the place where items get added, compared, and purchased. The platform that controls the transaction controls the game that actually pays. - [Today in AI — 23 May 2026](https://mitchellbryson.com/news/digest-2026-05-23): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The trillion-dollar proof of concept](https://mitchellbryson.com/news/trillion-dollar-proof-of-concept): OpenAI filed its trillion-dollar IPO on Friday. Days earlier, one of its models disproved an 80-year-old maths conjecture that amazed the field. Yet the filing follows a quarter with a negative 122% operating margin. The technology has never been more convincing; the business case has never been more precarious. - [AI just got a record deal](https://mitchellbryson.com/news/ai-just-got-a-record-deal): Spotify struck a licensing deal with Universal Music for fan-made AI covers and remixes, launched Studio as an AI podcast generator, and shipped an ElevenLabs-powered audiobook tool — while Stability AI released open-weight models capable of generating six-minute songs. The strategy is shifting from resistance to revenue. - [Today in AI — 22 May 2026](https://mitchellbryson.com/news/digest-2026-05-22): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 21 May 2026](https://mitchellbryson.com/news/digest-2026-05-21): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The lead changed during the commercial break](https://mitchellbryson.com/news/lead-changed-during-the-commercial-break): Four headlines in one week suggest Anthropic became the company the rest of the AI industry is reacting to. - [Today in AI — 20 May 2026](https://mitchellbryson.com/news/digest-2026-05-20): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The search box was the last passive interface](https://mitchellbryson.com/news/search-box-was-the-last-passive-interface): Google Search as you know it is over. - [Today in AI — 19 May 2026](https://mitchellbryson.com/news/digest-2026-05-19): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The chatbot ate the operating system](https://mitchellbryson.com/news/the-chatbot-ate-the-operating-system): Google is trying to make Gemini the interface that runs every screen and transaction. - [Today in AI — 18 May 2026](https://mitchellbryson.com/news/digest-2026-05-18): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Nobody budgeted for the humans](https://mitchellbryson.com/news/nobody-budgeted-for-the-humans): Samsung strike involving 47,000 workers looms as South Korea's president urges labor deal. - [Today in AI — 17 May 2026](https://mitchellbryson.com/news/digest-2026-05-17): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The narrowest boom in history](https://mitchellbryson.com/news/the-narrowest-boom-in-history): The haves and have nots of the AI gold rush. - [The best tool you're not allowed to keep](https://mitchellbryson.com/news/best-tool-youre-not-allowed-to-keep): Microsoft cancels Claude Code licences, shifting developers to GitHub Copilot CLI. - [Today in AI — 16 May 2026](https://mitchellbryson.com/news/digest-2026-05-16): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agents just moved in](https://mitchellbryson.com/news/agents-just-moved-in): Apple prepares App Store for autonomous AI agents. - [Today in AI — 15 May 2026](https://mitchellbryson.com/news/digest-2026-05-15): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 14 May 2026](https://mitchellbryson.com/news/digest-2026-05-14): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The AI arms dealer's dilemma](https://mitchellbryson.com/news/the-ai-arms-dealers-dilemma): Palo Alto Networks warns AI-driven cyberattacks will become the 'new norm' within months. - [Today in AI — 13 May 2026](https://mitchellbryson.com/news/digest-2026-05-13): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The swap that doesn't add up](https://mitchellbryson.com/news/the-swap-that-doesnt-add-up): AI-driven layoffs are failing to generate returns, landmark Gartner study finds. - [Today in AI — 12 May 2026](https://mitchellbryson.com/news/digest-2026-05-12): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The attack that wrote itself](https://mitchellbryson.com/news/the-attack-that-wrote-itself): Analysis of Google's interception of an AI-generated zero-day exploit and what divergent responses from OpenAI, Anthropic, and Microsoft mean for builders. - [Today in AI — 10 May 2026](https://mitchellbryson.com/news/digest-2026-05-10): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The chip king became the banker](https://mitchellbryson.com/news/the-chip-king-became-the-banker): Nvidia has committed over $40 billion in equity AI investments this year, financing everything from a $30 billion OpenAI stake to Corning's glass fibre plants ($3.2 billion). It is no longer just selling chips — it is bankrolling the entire stack that runs on them. Meanwhile, Anthropic's $1.8 billion deal with Akamai shows compute demand is so intense that even CDN companies are being recruited as AI infrastructure. The AI supply chain is being financially engineered by the chipmaker that profits at every layer. - [Today in AI — 9 May 2026](https://mitchellbryson.com/news/digest-2026-05-09): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The spreadsheet finally blinked](https://mitchellbryson.com/news/the-spreadsheet-finally-blinked): Three stories this week suggest the AI industry's financial narrative just hit its first real friction — not from regulators or competitors, but from the people doing the maths. SoftBank's lenders won't back OpenAI's private valuation at full price. Anthropic trades at $1 trillion on secondary markets where almost nobody is selling. And enterprises are burning 95 cents of every AI infrastructure dollar at 5% GPU utilisation. The trillion-dollar era has arrived, but the numbers are starting to argue with each other. - [They're cutting while they're winning](https://mitchellbryson.com/news/cutting-while-theyre-winning): In a single week, Cloudflare, DeepL, Freshworks, and Upwork all announced double-digit percentage workforce cuts — while posting revenue growth or beating expectations. The playbook is identical: rebrand as 'AI-native,' declare that smaller teams can do more with agents, and frame mass layoffs as transformation rather than cost-cutting. The AI restructuring wave has moved from early adopters to standard operating procedure, and it arrives when the numbers are good, not when they're bad. - [Today in AI — 8 May 2026](https://mitchellbryson.com/news/digest-2026-05-08): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 7 May 2026](https://mitchellbryson.com/news/digest-2026-05-07): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The night shift nobody asked for](https://mitchellbryson.com/news/the-night-shift-nobody-asked-for): Three announcements share a thread that should make builders take notice: AI that works when nobody's watching. Anthropic's 'dreaming' lets agents learn from their own mistakes between sessions, Claude Code Routines ship finished PRs while developers sleep, and Google DeepMind invested in the EVE Online studio to give AI a persistent virtual world for practicing long-term planning. The companies building always-on AI are betting the real value isn't in the minutes you use it — it's in the hours you don't. - [Today in AI — 6 May 2026](https://mitchellbryson.com/news/digest-2026-05-06): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agents got the job](https://mitchellbryson.com/news/the-agents-got-the-job): In a single 24-hour stretch, Anthropic shipped 10 preconfigured agents to Wall Street with Jamie Dimon's public blessing, Sierra raised $950 million because 40% of the Fortune 50 already relies on its customer service agents, PayPal announced it's cutting 4,500 jobs because AI can do the work, and a Y Combinator-backed startup handed the management of a real Stockholm cafe to an AI named Mona. The individual headlines say 'new tool' or 'new funding' — but the pattern says something bigger: AI agents crossed from experiment to employee this week, and nobody has the rulebook for what comes next. - [Today in AI — 5 May 2026](https://mitchellbryson.com/news/digest-2026-05-05): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The API wasn't enough](https://mitchellbryson.com/news/the-api-wasnt-enough): On the same day, OpenAI and Anthropic each launched PE-backed services companies that embed their own engineers inside enterprises. The shift from selling API access to selling implementation reveals something the labs have quietly concluded: the model alone doesn't close enterprise deals, and the fastest path to revenue runs through private equity's portfolio companies — not traditional sales cycles. - [Today in AI — 4 May 2026](https://mitchellbryson.com/news/digest-2026-05-04): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The presumption of AI](https://mitchellbryson.com/news/the-presumption-of-ai): VS Code stamps Copilot as co-author on commits even with AI disabled. OpenAI turns on ad tracking for free users by default. The Academy now demands proof that a performance is 'demonstrably human.' The industry crossed a line this week: AI involvement is assumed until you prove otherwise, and the burden of proof just shifted to everyone who makes things. - [Today in AI — 2 May 2026](https://mitchellbryson.com/news/digest-2026-05-02): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The one-model bet just died](https://mitchellbryson.com/news/one-model-bet-just-died): In one week, Microsoft shipped its most important AI product using competitors' models side by side, the Pentagon signed deals with seven AI providers instead of picking a winner, and xAI slashed prices by 40% to stay in the game. The lock-in playbook that defined the first AI era (one model, one provider, one platform) is dead. For anyone building on AI, the message is unambiguous: design for model portability, because your customers already expect it. - [Today in AI — 1 May 2026](https://mitchellbryson.com/news/digest-2026-05-01): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Not every AI dollar comes back](https://mitchellbryson.com/news/not-every-ai-dollar-comes-back): All four Big Tech hyperscalers reported earnings on the same night, collectively committing nearly $700 billion to AI infrastructure in 2026. But the market's response drew a sharp line: Alphabet, whose Google Cloud grew 63% on AI demand, was rewarded. Meta, which raised its AI capex to $145 billion, saw its stock drop 10% when Zuckerberg couldn't articulate the ROI. Meanwhile, Anthropic is racing to close a round that would value it above OpenAI at $900 billion. The industry's new question isn't 'are you investing in AI?' — it's 'what are you getting back?' - [Today in AI — 30 April 2026](https://mitchellbryson.com/news/digest-2026-04-30): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Nobody's locked in anymore](https://mitchellbryson.com/news/nobodys-locked-in-anymore): In the space of 48 hours, OpenAI broke free from Microsoft exclusivity to land on AWS, GM replaced Google Assistant with Gemini in 4 million cars, and Harvard dropped ChatGPT Edu in favour of Claude. The pattern is unmistakable: switching AI providers has become trivially easy at every level of the stack — cloud infrastructure, car dashboards, campus IT. For anyone building on top of AI, the implication is stark: your vendor relationship is a lease, not a mortgage, and the tenant can move out overnight. - [The biggest bet in tech just got called](https://mitchellbryson.com/news/biggest-bet-in-tech-just-got-called): OpenAI committed $600 billion in compute spending betting that revenue would keep doubling. This week, the Wall Street Journal revealed it missed its own growth targets and its CFO is publicly questioning the plan. Meanwhile, Google just poured $40 billion into Anthropic — whose revenue tripled in four months — and DeepSeek released a frontier-class model under MIT licence at a fraction of the cost. The first mover's advantage is starting to look like the first mover's burden. - [Today in AI — 29 April 2026](https://mitchellbryson.com/news/digest-2026-04-29): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 28 April 2026](https://mitchellbryson.com/news/digest-2026-04-28): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The model war is over — the device war just started](https://mitchellbryson.com/news/model-war-over-device-war-started): Three stories from the same week reveal the AI industry pivoting to its next phase. DeepSeek slashed model prices by 90%, confirming that raw intelligence is commoditizing fast. OpenAI responded not by cutting prices but by announcing plans to build its own smartphone where AI agents replace apps entirely. And Google poured $40 billion into Anthropic — buying not a model but the compute infrastructure underneath one. When the model becomes a commodity, the fight moves to who owns the device in your pocket and the datacenter behind it. - [Today in AI — 21 April 2026](https://mitchellbryson.com/news/digest-2026-04-21): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The latecomers just took the lead](https://mitchellbryson.com/news/latecomers-just-took-the-lead): Bezos builds a $38 billion AI lab in five months by poaching from everyone else. A young Beijing lab ships an open-weight model that beats Claude and GPT on coding benchmarks. Non-developers are flooding app stores with software they couldn't have built a year ago. The traditional advantages, years of research, massive teams, and first-mover status, are evaporating so fast that showing up late has become its own kind of edge. - [The AI labs are coming for your creative tools](https://mitchellbryson.com/news/ai-labs-coming-for-your-creative-tools): In a single week, Anthropic launched Claude Design (tanking Figma's stock 7%), Adobe countered with its agentic Firefly AI Assistant at Summit, and OpenAI expanded Codex with computer use, image generation, and 90+ plugins. The AI labs aren't just building models anymore — they're building the applications those models were supposed to power for someone else. For anyone in the design, creative, or developer tools business, the question is no longer whether the labs will compete with you, but how much runway you have left. - [Today in AI — 19 April 2026](https://mitchellbryson.com/news/digest-2026-04-19): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 18 April 2026](https://mitchellbryson.com/news/digest-2026-04-18): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The do-everything AI just became a liability](https://mitchellbryson.com/news/do-everything-ai-just-became-a-liability): OpenAI kills Sora and launches a specialist life sciences model in the same week, Cursor's coding-only bet hits a $50B valuation, and Novo Nordisk locks in domain-specific AI for drug discovery. The pattern is unmistakable: horizontal AI products are getting killed while vertical specialists attract all the money and attention. For anyone building on or with AI, the message is clear — pick a lane or get run over. - [The desktop just became the new front line](https://mitchellbryson.com/news/desktop-just-became-the-new-front-line): In a single 24-hour window, Google gave away a native Gemini Mac app, Perplexity began charging $200/month for an always-on desktop agent, and Mozilla dropped an open-source alternative for enterprises who refuse vendor lock-in. Three radically different business models, one shared conviction: the browser-tab era of AI was a transition phase, and the real interface war is for permanent residency on your desktop. - [Today in AI — 17 April 2026](https://mitchellbryson.com/news/digest-2026-04-17): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 16 April 2026](https://mitchellbryson.com/news/digest-2026-04-16): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The shortcut war just started](https://mitchellbryson.com/news/the-shortcut-war-just-started): Google launched Gemini for Mac with a system-wide Option+Space shortcut, Chrome turned prompts into reusable one-click Skills, and Microsoft gave Copilot the track-changes audit trail that lawyers demand. Three platform moves in 24 hours that share a thesis: the AI relationship will be won by whoever owns the surface area you already work on — the OS shortcut, the browser tab, the document. Standalone AI apps should be paying attention. - [Today in AI — 15 April 2026](https://mitchellbryson.com/news/digest-2026-04-15): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The labs are eating the app layer](https://mitchellbryson.com/news/labs-eating-the-app-layer): In one day, an unreleased Anthropic design tool wiped billions from Figma, Adobe, and Wix. OpenAI acqui-hired a personal finance startup to fold into ChatGPT. Microsoft revealed always-on autonomous agents for Copilot. The companies that built the models are now building the products — and every SaaS company just discovered its moat was always borrowed time. - [The confidence gap just got a scoreboard](https://mitchellbryson.com/news/confidence-gap-just-got-a-scoreboard): Stanford's annual AI Index and PwC's performance study both landed today with the same uncomfortable finding: there are two entirely different AI realities. Insiders see a gold rush — Vercel's riding agent-built apps toward an IPO, 'Claude mania' is sweeping conferences — while only 10% of Americans feel more excited than concerned about AI. The gap isn't just vibes: 74% of economic gains flow to 20% of companies, and Nature confirms humans still outperform AI agents on complex tasks. The optimism is earned, but so is the anxiety, and the numbers now prove both sides right. - [Today in AI — 14 April 2026](https://mitchellbryson.com/news/digest-2026-04-14): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 11 April 2026](https://mitchellbryson.com/news/digest-2026-04-11): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [They're fighting over your terminal now](https://mitchellbryson.com/news/fighting-over-your-terminal): OpenAI halved its Pro price to directly target Claude Code users. Anthropic shipped an advisor architecture that cuts agent costs by 85%. Shopify opened its platform to every major coding agent at once. The AI competition has moved from model benchmarks to the place where developers actually write code — and the pricing war that just started will reshape how every tool in the stack gets paid for. - [Today in AI — 9 April 2026](https://mitchellbryson.com/news/digest-2026-04-09): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Open source was always a phase](https://mitchellbryson.com/news/open-source-was-always-a-phase): Meta spent two years as AI's loudest open-source champion, building millions of developers into the Llama ecosystem and publishing manifestos about openness as a moral imperative. Today it shipped Muse Spark — its most competitive model yet — as fully proprietary, and the stock market rewarded the reversal with a 9% surge. For anyone who built their stack on the assumption that Big Tech open-source AI would stay open, this is the week to revisit that bet. - [Today in AI — 8 April 2026](https://mitchellbryson.com/news/digest-2026-04-08): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Too powerful to ship](https://mitchellbryson.com/news/too-powerful-to-ship): Anthropic built its most capable model and decided the world wasn't ready for it. A Chinese lab matched Western frontier performance on zero Nvidia hardware and gave it away under MIT license. Three rivals formed an unprecedented alliance to stop model copying. In a single day, the industry crossed a threshold: the models are now too powerful for their builders to ship freely, but too distributed to contain. For anyone building on top of these platforms, the question is no longer 'when will models be good enough?' — it's which capabilities your provider will actually let you use. - [Today in AI — 7 April 2026](https://mitchellbryson.com/news/digest-2026-04-07): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The labs are writing the rules now](https://mitchellbryson.com/news/the-labs-are-writing-the-rules-now): Anthropic's revenue tripled to $30 billion in four months; on the same day, OpenAI published a 13-page economic blueprint proposing robot taxes, public wealth funds, and a four-day workweek. These aren't separate stories — they're a single inflection point. When companies generate this much revenue this fast, they stop lobbying for policy and start drafting the economic framework for the world they're reshaping. Japan, already deploying physical AI to fill jobs nobody else wants, is living the preview. - [Today in AI — 6 April 2026](https://mitchellbryson.com/news/digest-2026-04-06): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The fine print is the product now](https://mitchellbryson.com/news/the-fine-print-is-the-product): This week's biggest stories share a common thread: the implicit deal between AI tools and their users is being renegotiated — and the outcomes diverge wildly depending on who shows up to the table. Hollywood writers secured expanded AI protections through proactive negotiation, while GitHub quietly changed Copilot's terms to train on developers' code by default. For anyone building with or on top of AI, the lesson is the same: the fine print now matters more than the feature set. - [Today in AI — 5 April 2026](https://mitchellbryson.com/news/digest-2026-04-05): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Headcount just became a vanity metric](https://mitchellbryson.com/news/headcount-just-became-a-vanity-metric): Both Sam Altman and Dario Amodei predicted a billion-dollar solo-founder company would arrive in 2026. Medvi just proved them right — $1.8B tracking revenue, two employees, AI handling everything from code to customer service. Meanwhile, Anthropic shipped a multi-agent harness that autonomously builds full-stack apps in 4-hour sessions, PrismML put competitive LLMs on a phone, and Utah let an AI chatbot renew psychiatric prescriptions. The minimum viable team for serious economic output is converging on one. - [Today in AI — 4 April 2026](https://mitchellbryson.com/news/digest-2026-04-04): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The handshake era is over](https://mitchellbryson.com/news/the-handshake-era-is-over): Within 48 hours, the three biggest AI companies each made a move to own more of their stack. Anthropic blocked third-party tools from piggybacking on Claude subscriptions. Microsoft launched in-house foundation models to reduce its dependence on OpenAI. OpenAI bought a media company to control its own narrative. The mix-and-match era of AI — where you could freely combine models, tools, and distribution from different providers — is giving way to vertical integration and platform control. - [The AI giants are building empires, not products](https://mitchellbryson.com/news/ai-giants-building-empires-not-products): OpenAI buys a media company, launches self-serve ads, and cuts prices to lock in developers. Microsoft ships its own models to compete with the partner it funds. Google opens Gemini's doors and offers to import your ChatGPT history. These aren't product announcements — they're land grabs for distribution, narrative control, and platform lock-in. The race has shifted from who has the best model to who owns the surface area around it. - [Today in AI — 3 April 2026](https://mitchellbryson.com/news/digest-2026-04-03): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 2 April 2026](https://mitchellbryson.com/news/digest-2026-04-02): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The conversation is the storefront](https://mitchellbryson.com/news/the-conversation-is-the-storefront): In the span of weeks, AI conversations went from being tools to being storefronts. ChatGPT hit $100M in ad revenue and is opening self-serve access to any business this month. Gap enabled direct checkout inside Gemini — you can buy jeans without leaving the chat. And Perplexity got sued for allegedly sharing private conversations with ad networks all along. The AI industry's business model is crystallizing, and it looks a lot like the internet's original sin — except this time the data is far more intimate than a search query ever was. - [Today in AI — 1 April 2026](https://mitchellbryson.com/news/digest-2026-04-01): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The most expensive code nobody meant to ship](https://mitchellbryson.com/news/the-most-expensive-code-nobody-meant-to-ship): The same week the AI industry asked investors for its biggest vote of confidence, accidental source leaks, supply chain breaches, and mass layoffs revealed an execution gap the market has not yet priced in. - [Today in AI — 31 March 2026](https://mitchellbryson.com/news/digest-2026-03-31): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Models are commodities, platforms are the prize](https://mitchellbryson.com/news/models-are-commodities-platforms-are-the-prize): Microsoft pairs GPT and Claude as interchangeable components in a single workflow. Meta considers licensing Gemini because its own model can't keep up. Apple opens Siri to every chatbot via an Extensions marketplace. The pattern is unmistakable: the companies that sit between users and models — not the labs that train them — are positioning themselves as the durable layer. The frontier labs are becoming suppliers in someone else's platform play, and the IPO race between Anthropic and OpenAI may be as much about locking in platform status as it is about revenue. - [Today in AI — 30 March 2026](https://mitchellbryson.com/news/digest-2026-03-30): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [OpenAI just showed everyone what an AI company looks like when the math stops working](https://mitchellbryson.com/news/openai-math-stops-working): In a single week, OpenAI killed Sora ($15M/day burn, $2.1M lifetime revenue), blindsided Disney on a $1B deal, shelved its adult chatbot, renamed its product org to 'AGI Deployment,' moved safety oversight away from the CEO, and bet everything on a model codenamed 'Spud.' This isn't a pivot — it's a company strip-mining itself to win a race it can't afford to lose, and the collateral damage reveals how thin the margin for error is when you're spending faster than you're earning. - [Today in AI — 29 March 2026](https://mitchellbryson.com/news/digest-2026-03-29): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The real AI war is over your memory](https://mitchellbryson.com/news/the-real-ai-war-is-over-your-memory): Google launched tools to import your ChatGPT memories and chat histories. Apple is turning Siri into a marketplace where every AI assistant plugs in — for a 30% cut. OpenAI is wiring Codex into every work tool you touch. Shopify made every AI conversation a storefront. The model quality race is giving way to something more consequential: a fight over who accumulates the deepest context about how you work, shop, and think — because whoever owns that context makes switching nearly impossible. - [Today in AI — 28 March 2026](https://mitchellbryson.com/news/digest-2026-03-28): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The leak that repriced cybersecurity](https://mitchellbryson.com/news/the-leak-that-repriced-cybersecurity): Anthropic's accidental Mythos reveal crashed cybersecurity stocks — but the market was catching up to a reality that was already here. On the same day, CISA warned of active exploitation of AI agent frameworks, researchers disclosed basic vulnerabilities in LangChain's widely-used tooling, and IBM confirmed the first AI-generated malware in a live ransomware campaign. AI is simultaneously the most powerful offensive tool, the most attacked infrastructure, and the newest malware author — and the market just priced that triple threat in. - [Today in AI — 27 March 2026](https://mitchellbryson.com/news/digest-2026-03-27): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The text box was just the prototype](https://mitchellbryson.com/news/text-box-was-just-the-prototype): In a single 48-hour stretch, Apple revealed plans to open Siri to rival chatbots, Mistral shipped an open-source voice model rivaling ElevenLabs, Google launched studio-quality AI music generation, and IBM embedded voice AI into enterprise agents. The chat window got AI onto our screens — voice and audio are how it escapes into the rest of our lives. The companies that dominate the next phase of AI won't be the ones with the best text completions; they'll be the ones that sound the most human. - [The checkout wars have begun and nobody agrees on the rules](https://mitchellbryson.com/news/checkout-wars-nobody-agrees-on-the-rules): In a single week, OpenAI killed Instant Checkout and pivoted to visual discovery, Gap became the first major retailer to let you buy inside Gemini, Shopify opened its Universal Commerce Protocol to every brand, and Stripe shipped a protocol for agents to spend money without humans. The AI shopping land grab is happening in real time — but the industry is split on the most fundamental question: should you buy inside the AI, or should the AI send you back to the store? - [Today in AI — 26 March 2026](https://mitchellbryson.com/news/digest-2026-03-26): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 25 March 2026](https://mitchellbryson.com/news/digest-2026-03-25): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The toy phase is over](https://mitchellbryson.com/news/toy-phase-is-over): OpenAI killed Sora and let Disney's billion-dollar investment walk away — the flashiest product in AI, sacrificed so the company can focus on productivity tools ahead of its IPO. On the same day, Anthropic shipped desktop computer use so Claude can literally control your Mac while you step away, and Arm unveiled the first chip it has ever manufactured — a 136-core processor purpose-built for the agent data center. The spectacle era is being traded for the things that actually ship, sell, and scale. - [Today in AI — 24 March 2026](https://mitchellbryson.com/news/digest-2026-03-24): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The land grab has gone financial](https://mitchellbryson.com/news/the-land-grab-has-gone-financial): OpenAI's 17.5% guaranteed-return PE pitch, its 450,000 sq ft campus lease, and the Helion fusion deal all point to the same shift: the AI race is no longer about who has the best model — it's about who can lock in distribution, real estate, and energy before the IPO window opens. Anthropic is running the same playbook at smaller scale, and Musk is vertically integrating chips. The competition has moved from the lab to the balance sheet, and the winners will be decided by deal structure as much as by benchmark scores. - [Compute is the new equity](https://mitchellbryson.com/news/compute-is-the-new-equity): In a single week, Nvidia proposed paying engineers half their salary in AI inference tokens, OpenAI announced plans to nearly double its headcount to 8,000, and data showed Anthropic capturing 73% of new enterprise AI spending with a fraction of the workforce. The AI industry is splitting into three competing theories of value: raw compute, human talent, and enterprise relationships, and Jensen Huang's token paycheck is the clearest signal yet that the boundaries between them are collapsing. When your raise comes in tokens and your productivity is measured in inference, the employment contract is being rewritten in real time. - [Today in AI — 23 March 2026](https://mitchellbryson.com/news/digest-2026-03-23): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 21 March 2026](https://mitchellbryson.com/news/digest-2026-03-21): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The superapp moment has arrived](https://mitchellbryson.com/news/the-superapp-moment-has-arrived): OpenAI's Fidji Simo told employees the company was 'spreading efforts across too many apps.' Google is sunsetting Firebase Studio to fold everything into AI Studio. Meanwhile, Carta data shows the same consolidation logic playing out in capital markets: 10% of AI startups now capture half of all venture funding. After two years of tool proliferation, the platforms have decided the winner isn't the best model — it's the best starting screen. - [Today in AI — 20 March 2026](https://mitchellbryson.com/news/digest-2026-03-20): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The workflow is the new moat](https://mitchellbryson.com/news/workflow-is-the-new-moat): Three stories from the same day reveal that AI companies have collectively pivoted from competing on model benchmarks to competing on where people actually work. OpenAI acquired Astral — a company that makes linters and package managers, not AI — to embed itself deeper into coding workflows. Google turned Docs and Sheets into Gemini co-authoring surfaces. And ChatGPT gained write actions inside Outlook and Google Apps, shifting from advisor to actor. The model is becoming a commodity; the integration point is becoming the moat. - [Today in AI — 19 March 2026](https://mitchellbryson.com/news/digest-2026-03-19): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The smallest models are carrying the biggest bets](https://mitchellbryson.com/news/smallest-models-biggest-bets): OpenAI, Mistral, and Nvidia all shipped their smallest models this week — and far from being a concession, these tiny models are the linchpin of the agent revolution. GPT-5.4 nano costs pennies per million tokens. Mistral Small 4 activates just 6B of its 119B parameters per query. Nvidia rallied eight AI labs to build open models that run on desktop hardware. The industry is collectively betting that the next phase of AI depends not on building bigger brains, but on making intelligence cheap and fast enough to run everywhere, all the time, which is exactly what the agent era requires. - [The always-on agent needs its own computer](https://mitchellbryson.com/news/always-on-agent-needs-its-own-computer): In a single week, the AI industry acknowledged that agents are a fundamentally different computing paradigm. Perplexity shipped a $200/month always-on agent running on dedicated Mac mini hardware. Microsoft built Copilot Cowork as a persistent cloud agent with access to your entire enterprise graph. And Nvidia is redesigning its CPUs specifically for agentic workloads, with Jensen Huang calling them 'the bottleneck.' The chat-window era is ending — agents need persistent infrastructure at every layer of the stack, and builders who treat them as fancy API wrappers are building for yesterday. - [Today in AI — 14 March 2026](https://mitchellbryson.com/news/digest-2026-03-14): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The agent stack just got its operating system](https://mitchellbryson.com/news/agent-stack-gets-its-operating-system): In a single week, every layer of the AI agent stack advanced simultaneously: Microsoft shipped Agent 365 as an enterprise control plane for governing fleets of AI agents, Google open-sourced ADK for TypeScript so web developers can build multi-agent systems, Meta acquired a social network where agents talk to each other, and Mandiant's founder raised $190M to build autonomous security agents. For product builders, the signal is unmistakable — agents have crossed from prototype to platform, and the race to define how they're built, deployed, managed, and secured is happening right now. - [Today in AI — 12 March 2026](https://mitchellbryson.com/news/digest-2026-03-12): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 11 March 2026](https://mitchellbryson.com/news/digest-2026-03-11): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Everyone is building the agent operating system](https://mitchellbryson.com/news/everyone-building-agent-operating-system): In a single 48-hour window, Microsoft shipped Agent 365 (the 'control plane' at $15/user/month), NVIDIA announced NemoClaw (an open-source, hardware-agnostic agent orchestration platform), Meta acquired the social network where AI agents already talk to each other, and Google deployed an agent-builder for 3 million Pentagon users. The model race has quietly given way to a platform race: who gets to observe, govern, route, and monetize the AI agents flooding every enterprise? For builders, this is the moment the agent infrastructure layer starts to crystallize — and the platform you pick now may determine what you're locked into for years. - [AI finally learns to secure the code it writes](https://mitchellbryson.com/news/ai-learns-to-secure-the-code-it-writes): OpenAI shipping Codex Security, Anthropic's Claude finding 22 CVEs in Firefox in two weeks, and Microsoft treating AI agents as governed security principals all point to the same inflection: the industry is racing to close the security gap that AI coding tools opened. With research showing only 10% of AI-generated code is both functional and secure, the companies that created the problem are now building the remediation layer — and the ones who move fastest will own the trust infrastructure for the agent era. - [Today in AI — 10 March 2026](https://mitchellbryson.com/news/digest-2026-03-10): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [AI's most important new trick is knowing when to stop thinking](https://mitchellbryson.com/news/ai-efficiency-era-knowing-when-not-to-think): Three separate teams shipped models this week that cut token usage by nearly half through the same insight: the best inference is often no inference at all. - [Today in AI — 7 March 2026](https://mitchellbryson.com/news/digest-2026-03-07): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 5 March 2026](https://mitchellbryson.com/news/digest-2026-03-05): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Three models, three bets on what intelligence becomes](https://mitchellbryson.com/news/three-models-three-bets-on-intelligence): OpenAI, DeepSeek, and Alibaba each shipped new models this week, and each made a fundamentally different bet about where AI capability needs to go next. - [Today in AI — 4 March 2026](https://mitchellbryson.com/news/digest-2026-03-04): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The security bill is coming due for AI's agent era](https://mitchellbryson.com/news/security-bill-ai-agent-era): AI agents are gaining deeper access to enterprise systems and developer environments faster than anyone is securing them. Three stories from a single news cycle show the attack surface widening in real time. - [Today in AI — 3 March 2026](https://mitchellbryson.com/news/digest-2026-03-03): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The Pentagon values auction: AI safety gets its market test](https://mitchellbryson.com/news/pentagon-values-auction-ai-safety-market-test): OpenAI amends its Pentagon deal after Altman admits it looked 'opportunistic and sloppy', while Claude surges to number one on the App Store and hundreds of employees publicly back Anthropic's stance. - [The agent security reckoning nobody is ready for](https://mitchellbryson.com/news/agent-security-reckoning): Three separate security disclosures this week exposed a pattern: we are deploying agentic AI infrastructure faster than we can secure it, from MCP servers to coding assistants. - [Today in AI — 2 March 2026](https://mitchellbryson.com/news/digest-2026-03-02): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [Today in AI — 1 March 2026](https://mitchellbryson.com/news/digest-2026-03-01): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [The infrastructure tax comes due](https://mitchellbryson.com/news/infrastructure-tax-comes-due): Hyperscalers committed $700 billion to AI data centres in 2026 while NVIDIA posted record revenue and memory chip shortages drive up consumer device prices globally. - [Today in AI — 28 February 2026](https://mitchellbryson.com/news/digest-2026-02-28): Today's top AI news — curated links and commentary on the stories that matter for product builders. - [OpenAI's $110B round is really an infrastructure lock-in deal](https://mitchellbryson.com/news/openai-110b-funding-infrastructure-lock-in): Amazon's $50 billion investment buys exclusive cloud hosting for OpenAI's enterprise platform, creating a split-brain infrastructure that every builder on OpenAI needs to understand. - [Trump Administration Orders Federal Agencies to Cease Business with Anthropic Over Pentagon Restrictions](https://mitchellbryson.com/news/trump-bans-anthropic-pentagon-ai-restrictions): The Trump administration ordered federal agencies and military contractors to stop working with Anthropic after the company refused to allow unrestricted military use of Claude AI, including for autonomous weapons and mass surveillance. - [AI News Roundup — February 27, 2026](https://mitchellbryson.com/news/digest-2026-02-27): This week's top AI news, product launches, and trending repos - [Google DeepMind Unveils Nano Banana 2: Revolutionary AI Image Generator Democratizes Professional Creation](https://mitchellbryson.com/news/google-nano-banana-2-revolutionary-ai-image-generator): Google DeepMind launched Nano Banana 2 (Gemini 3.1 Flash Image), blending high-quality outputs with unprecedented speed to democratize professional-grade image creation across Google's product suite. - [Pentagon Escalates Dispute with Anthropic, Threatens Defense Production Act](https://mitchellbryson.com/news/pentagon-anthropic-ai-dispute-escalates): Defense Secretary Pete Hegseth gives Anthropic until Friday to provide military access to Claude or face being declared a supply chain risk or forced compliance under the Defense Production Act. - [AI News Roundup — February 21, 2026](https://mitchellbryson.com/news/digest-2026-02-21): This week's top AI news: the agentic stack matures with MCP standardisation and new orchestration platforms, Hollywood escalates the copyright fight, and a delicious meta-irony in AI journalism. - [Observational Memory Might Kill RAG for Long-Running Agents](https://mitchellbryson.com/news/observational-memory-agents): Mastra's open-source memory system replaces retrieval with background compression, scoring highest on LongMemEval at a fraction of the cost. If you're building agents that need to remember, pay attention. ## Uses - [Claude Code](https://mitchellbryson.com/uses/claude-code): My daily driver for AI-assisted coding and running agentic teams. Claude works directly in the codebase via terminal, IDE, or web - great for multi-step tasks and autonomous workflows. - [Cursor](https://mitchellbryson.com/uses/cursor): AI-powered IDE built on VS Code. I use it mainly for codebase exploration - understanding unfamiliar repos, navigating large codebases, and quick edits with strong context awareness. - [Codex](https://mitchellbryson.com/uses/codex): OpenAI's coding agent for background tasks like issue triage, CI monitoring, and automated workflows. Runs headless or via CLI without blocking the main workflow. - [GitLens](https://mitchellbryson.com/uses/gitlens): Git superpowers for VS Code and Cursor. Inline blame, commit graph, worktrees, and AI commit messages. Makes Git history and collaboration much easier. - [n8n](https://mitchellbryson.com/uses/n8n): Self-hostable workflow automation. Connects APIs, databases, and services with a visual editor. I use it for data pipelines, webhooks, and glue between tools. - [Metabase](https://mitchellbryson.com/uses/metabase): Open-source BI and data visualisation. Build dashboards, run ad-hoc queries, and share insights without writing SQL. Works well with Postgres, MySQL, and more. - [Supabase](https://mitchellbryson.com/uses/supabase): Hosted Postgres with auth, storage, and realtime. Open-source Firebase alternative. I use it for backend-as-a-service and vector search via pgvector when I need a database without managing servers. - [Upstash](https://mitchellbryson.com/uses/upstash): Serverless Redis with per-request pricing. No connection limits, works at the edge. I use it for caching, rate limiting, vector search, and session storage in serverless apps. - [Wispr Flow](https://mitchellbryson.com/uses/wispr-flow): Voice-to-text that actually works for code. Handles camelCase, acronyms, and dev jargon. I use it for docs, notes, and long-form writing when typing is slow. - [Obsidian](https://mitchellbryson.com/uses/obsidian): Markdown-based note-taking with local-first storage and a graph view. My second brain for projects, meeting notes, and knowledge bases that link together. - [Warp](https://mitchellbryson.com/uses/warp): Modern terminal with AI, blocks, and a polished UX. Better autocomplete, command palette, and split panes. Replaced iTerm as my default shell. - [Rectangle](https://mitchellbryson.com/uses/rectangle): Open-source Mac window manager. Keyboard shortcuts and snap zones to tile, maximise, and arrange windows. Replaced Spectacle for me - free and actively maintained. - [DigitalOcean](https://mitchellbryson.com/uses/digitalocean): Cloud hosting for VPS, managed databases, and app platforms. Simple pricing, good docs, and Droplets that scale. I use it for side projects and self-hosted services. - [Vercel](https://mitchellbryson.com/uses/vercel): Frontend and serverless hosting optimised for Next.js. Zero-config deploys, edge functions, and preview URLs per branch. My go-to for static sites and JAMstack apps. - [Dokku](https://mitchellbryson.com/uses/dokku): Self-hosted Heroku-style PaaS. Git push to deploy, buildpacks, and one-command app management. Runs on any VPS - great for small teams who want control without Kubernetes. - [Cloudflare](https://mitchellbryson.com/uses/cloudflare): DNS, CDN, and edge security. Fast propagation, free tier, and DDoS protection. I use it for domain management, SSL, and sometimes Workers for lightweight edge logic. - [Namecheap](https://mitchellbryson.com/uses/namecheap): Domain registrar with competitive pricing and straightforward management. I use it to register and renew domains, then point them at Cloudflare or hosting. - [Resend](https://mitchellbryson.com/uses/resend): Transactional email API built for developers. Simple SDK, React Email support, and good deliverability. Powers contact forms, notifications, and password resets.