62 posts across Systems & Adoption, Supervising Agents, Directing AI, Work & Judgement, Judging AI
McKinsey found 94% of companies still see no measurable AI return despite licenses everywhere. Four separate 2026 moves — a monetized free tier, outcome-based consulting fees, and a user base that's no longer mostly developers — show what actually separates the 6% that do.
5 min readOpenAI's own researchers now run more agent-workdays than human ones. Watching a second and third agent isn't a bigger version of watching your first — it's a different problem, and most people are solving the wrong one.
5 min readSix separate product launches since April, from always-on Gmail agents to no-code voice builders, kept landing on the same finding. Writing a better prompt was never the fix — defining scope before you delegate is.
5 min readFour separate AI launches this year taught the same lesson from different angles: model choice matters less than task fit. Here are the three questions to ask before you pick a model for anything that matters.
4 min readAnthropic and OpenAI shipped their best models within two days of each other, and independent benchmarks show them trading wins. The differentiator left standing is how well you evaluate and supervise either one.
3 min readEthan Mollick's research found the models best at doing your work alone score lower on helping you think it through, while a cheaper model does both well. Here's the three-question check for matching model to task.
4 min readThe scope and guardrails you built for one agent don't cover what happens when it hands off to a second agent instead of to you. Here's what changes when agents work together.
6 min readOpenAI, NEC, and Shopify have each committed real money and structure this year to teaching people how to use AI — not to making the model smarter. Five signals from 2026 point at the same bottleneck: adoption, not access.
5 min readA silent 40% token hike, a subsidized subscription, a budget built on sticker price, and an agent deployed where a prompt would do — four unrelated cost overruns from this year share one root cause. Here's the check that catches it before the invoice does.
5 min readOne recent agent incident shows a pattern most guardrail advice doesn't cover: the goal was scoped correctly and the permissions were appropriate, and the agent still did something nobody would have approved. Here's the constraint that catches what scoping and permissions don't.
5 min readAnthropic moved computer use, browser use, the Files API, and the Skills API to general availability on August 19 — meaning production SLAs, not beta disclaimers. That changes what you need to decide before turning one loose on real work.
4 min readClaude, Gemini, and DeepSeek all shipped the same change this year — explicit control over how hard the model thinks, priced across a 50x range. Here's the three-tier rule for matching effort to what a task actually needs.
4 min readGoogle's new Sheets Canvas feature turns a plain-English prompt into a working app on top of your live spreadsheet data. Here's what changes for anyone who runs a business process out of a spreadsheet, and what the prompt still has to specify.
4 min readA leadership exodus, a 19-day suspension, a government approval gate, and a pricing model that can't hold — four separate incidents this year, four different mechanisms, one lesson: the AI vendor you built around was never a fixed point.
5 min readSix product launches since April removed the moment where you used to check AI output before it acted. None of them shipped a replacement — here's how to build one back in.
4 min readDeepSeek's new model costs a fraction of a cent per million tokens, and OpenAI just cut its own prices 80% in three weeks. The tasks got cheaper — knowing which tier each one needs didn't.
4 min readGoogle's Gemini Spark can now auto-browse the web using your real logins and saved passwords — and it keeps running after you close your laptop. Three boundaries to set before you switch it on.
4 min readPublic sector AI use carries constraints the private sector doesn't: decisions must be explainable, records must be defensible, and the people affected can't take their custom elsewhere. Here's what that changes.
5 min readEthan Mollick's Summer 2026 AI guide marks the defining shift: we've moved from tools that answer to tools that act. The skill that matters now isn't how well you prompt — it's knowing what your AI is allowed to do without asking.
5 min readLenny Rachitsky's 6,000-person survey found the workforce splitting along an AI skill fault line — not access, not tools, but whether you've built the workflows that let AI amplify rather than overwhelm. Here's what the amplified 49% are doing differently.
5 min readAnthropic shipped an effort toggle with Claude Opus 5. Google released three tiers of Gemini Flash and deprecated temperature settings. Every major AI provider made the same call this week: match the task to the reasoning depth. Here's how.
5 min readProject management is mostly synthesis, chasing, and translation between audiences — three things AI handles well. Here's how to hand those over without losing the judgement that keeps a project honest.
5 min readMost people get a genuinely good AI result, feel pleased, and then start from scratch the following week. The gap between occasional wins and compounding ones is about twenty minutes of writing.
5 min readMcKinsey reports 30%+ of its consulting fees are now tied to client outcomes, not hours. Deloitte warns labour-based consulting could shrink significantly by 2035. Here's what the shift from hours to outcomes means for every professional who sells time.
5 min readCourt documents unsealed July 2 showed Anthropic drew two non-negotiable redlines — no mass surveillance, no autonomous weapons — and lost government access because of it. The 19-day Fable 5 suspension was the operational consequence. Here's what it means for how you build with AI.
5 min readGoogle shipped desktop task automation via your phone. xAI launched a no-code voice agent builder at $0.05/minute. The tools that used to require a developer are now for everyone — which means the skills required to use them safely just became yours to learn.
6 min readAgents are the wrong shape for more tasks than the enthusiasm suggests — and for a few of them, an old-fashioned rule or a person is better. Six signals that the answer isn't an agent.
5 min readA study of 18,978 conversations found AI nearly three times more effective at fundraising than trained human canvassers — but only when unconstrained. Match the speed and length, and the advantage disappears. The practical implication for your writing workflow is not what you'd expect.
5 min readBoth OpenAI and Anthropic have their best models gated by government approval — not pricing tiers, not waitlists. If you built workflows around those models, here are three things to do before it happens again.
5 min readFive independent signals converged in one week on the same conclusion: the browser-tab AI model is ending. Mollick's Codex research, Karpathy on Claude Tag, and Anthropic's own numbers tell you what to practise next.
5 min readA NEJM AI study of 376 unresolved cases shows what the best AI collaboration pattern looks like in practice: AI generates hypotheses, humans decide what to do with them. Here is the structure worth applying to any analysis-heavy work.
4 min readOn June 16, OpenAI launched its Partner Network with $150M and a December target of 300,000 certified AI consultants, backed by Accenture, BCG, McKinsey, and PwC. What professionalisation means for professionals who have been learning on their own.
5 min readOn June 17, Altman, Amodei, and Hassabis sat together at the G7 summit and gave their clearest AGI timeline estimates yet. What that means for professionals still treating AI skill-building as optional.
5 min readOn June 18, Google DeepMind published a framework that classifies its own autonomous agents as insider threats requiring defense-in-depth monitoring. What that means for every team now deploying AI agents.
5 min readUber exhausted its 2026 AI budget in four months. Sam Altman says enterprise cost complaints are now 'a meme.' The root cause isn't reckless spend — it's deploying always-on agents for tasks a single prompt could handle. Here's the three-tier framework that fixes it.
5 min readAlmost every team that adopts AI builds a shared prompt library, and almost all of them are abandoned within a quarter. The reasons are consistent, and so is the version that survives.
5 min readDario Amodei's 15,000-word essay said AI may cause permanent job displacement. The press release that followed the same day created 1,000 paid roles for people who can help organizations adopt it. The skills those roles need are learnable now.
4 min readThree product shifts in five days — Mollick's 'patron, not wizard' frame, OpenAI's reasoning tiers, and Claude's scheduled agents — prove the professional AI skill stack has moved on from prompting.
6 min readAnthropic revealed that AI writes 80% of its own code — and called for a global pause option in case something goes wrong at scale. If you're using AI for any workflow output that matters, you already face the same question: how do you maintain oversight when volume outpaces review?
5 min readOpenAI's Codex hit 5M weekly users — and 1 in 5 is a non-technical professional using it for data analytics, sales, and product design. In the same week, ChatGPT memory jumped from 41.5% to 82.8% accurate. The skill gap is no longer about code.
5 min readUber blew its annual AI budget in months. Simon Willison's $200 subscription runs $2,180 in compute. Anthropic's revenue went 5x in five months. If your team budgeted for AI tools in Q4, those numbers are already wrong — here's how to reset them.
4 min readClaude Opus 4.8 shipped with explicit effort controls — users choose whether the model reasons carefully or answers fast, at 3x the cost difference. Here's how to decide which tasks deserve which mode.
4 min readClaude Pro subscribers consume $200–400 of compute per $20/month. OpenAI loses $1.22 for every $1 earned. GitHub Copilot switches to usage-based billing June 1. The price you pay for AI today is not the price it costs to run. Here's how to plan for when that changes.
5 min readEnterprise AI rollout advice is written for organisations with a change function. For a team of ten, most of it is overhead. Here's the sequence that works at small scale.
6 min readJosh Comeau watched an expert developer triple their output with AI and watched beginners spend three hours prompting a model to solve something they fixed manually in 30 seconds. The difference wasn't the model. It was what each person brought to it.
4 min readA prompt asks for output you'll read next. A brief defines work that runs without you. The difference is four things a prompt never has to specify — and skipping them is why most agent deployments quietly fail.
5 min readTwo Hacker News posts hit the front page the same week with a combined 1,374 points. One named the upstream problem — vague briefs going in. One named the downstream problem — raw output going out. The model was fine in both cases.
4 min readGoogle launched Gemini Spark at I/O 2026 — a 24/7 personal agent that runs inside Gmail, Docs, and Calendar, acting even when your device is off. The question is no longer whether to use AI at work. It's what the agent is doing right now, and whether you authorised it.
5 min readGoogle's AI Pointer and Gemini Rambler let you point, speak, and get results without typing a word. The syntax of prompting is being absorbed into the interface. The judgment behind it isn't.
4 min readShopify runs its internal AI coding agent exclusively in public Slack channels — never DMs. The whole company learns by watching. Here's how to apply the same principle without building a custom agent.
3 min readOpenAI is spending $4B to send consultants into organizations to teach them how to use AI. Malta requires an AI literacy course before citizens get free ChatGPT Plus. Both moves prove the same thing: the bottleneck isn't access to the tools — it's knowing what to do with them.
4 min readAnthropic's CEO spent most of 2025 warning AI could eliminate 50% of entry-level knowledge work. On May 5 he invoked the Jevons Paradox instead. When the person building the tool revises his own worst-case prediction, the signal isn't reassurance — it's a timestamp on a window that's still open.
5 min readOpenAI's B2B Signals data shows the AI gap has stopped being about who uses it and started being about what they use it for. Typical firms ask AI questions. Frontier firms build AI into multi-step execution — and the difference is widening.
5 min readSimon Willison admitted he skips reviewing AI code for production systems. A Stockholm cafe let AI order 120 eggs for a kitchen with no stove. Ethan Mollick says we don't yet have words for how multi-agent systems fail. The pattern is the same: as AI gets more capable, the temptation to stop checking grows — and the cost of not checking doesn't shrink.
5 min readGoogle's Gemini can now output a finished Word file, spreadsheet, or PDF directly from conversation. Writing prompts that produce share-ready documents is a different skill than writing prompts that produce text.
4 min readOpenAI shipped image generation that produces up to 8 consistent, text-accurate visuals from a single prompt. It works. The gap between that capability and campaign-ready assets comes down to one thing: the quality of your visual brief.
4 min readOpenAI and Google shipped always-on agent platforms in the same week. Non-technical teams can now set up agents by describing workflows in plain English. The skill gap is no longer how to prompt — it's how to scope, delegate, and verify autonomous work.
5 min readApproving everything forever doesn't scale, and switching an agent to autonomous on day one is how the incident stories start. The workable pattern is a ramp, and it looks a lot like onboarding a new colleague.
5 min readWhen Anthropic updated Claude to Opus 4.7, real-world token usage jumped roughly 40% — with no pricing announcement. Here's what that means for any team running AI workflows at any scale.
5 min readAndrej Karpathy published a system that turns a folder of raw notes into a living, self-updating wiki — without writing a word of it himself. The practical version is closer than you think.
5 min readAnthropic's Cowork, Perplexity's Comet, and ChatGPT in CarPlay all launched within weeks. AI agents are no longer a developer concept. Here's how to actually use them.
5 min readOpenAI killed Sora and redirected compute toward automated researchers. That's a capital allocation signal about which knowledge work skills survive. Three specific ones to invest in now.
5 min readWe use one cookie to keep you signed in — that one isn't optional. Separately, we'd like to record which drills people finish and where they get stuck, so we can fix the parts that don't teach. No ad networks, no selling anything on. Privacy policy.