27 posts across Directing AI, Judging AI, Supervising Agents
Three product changes this year quietly removed the editing pass between AI output and the finished file. Here's what has to move into the prompt now that nothing catches a gap afterward.
4 min readAn iPhone update, a developer-productivity report, a persuasion study, and an AI-detection tool landed within six weeks of each other this summer. None of them were about prompting technique — and that's the point worth noticing.
4 min readGPT-6 and Claude both started choosing their own output format this week — dashboards, calculators, animations — instead of defaulting to prose. The model deciding what you see is a new layer to check before you trust it.
4 min readFour separate stories this year — a CEO's warning, a viral study, a blunt writing rule, a $50,000 gap in financial advice — all point at the same three seams. Here's the throughline, and the move that closes each one.
4 min readSimon Willison endorsed a rule blunt enough to actually follow: write it yourself, then use AI only to flag problems, never to supply the fix. Two studies published the same week explain why the line matters more than it sounds.
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 readSubstack launched AI-detection scanning on July 21, powered by Pangram. Any reader can now check any post over 100 words for human vs AI content — one click, no subscription required. Content authenticity is now a visible, reader-facing metric. Here's what to change before you publish again.
5 min readThe useful question isn't whether writers should use AI. It's which stage of writing to hand over — because handing over the wrong one costs you the thing people are paying for.
5 min readOpenAI tripled the custom instructions limit for paid users on July 15. The people who use this well will stop re-explaining themselves every conversation. Here's what a useful briefing document actually contains.
4 min readCursor's July 2026 data shows the top 1% of AI users produce 46 times more than the median. Everyone's using the same tools. The gap is a skill — and it's learnable.
5 min readOpenAI's own safety guidance for ChatGPT Work admits the system can produce 'finished mistakes at a scale that is harder to catch.' The same week, Wharton research killed prompt tricks and Cursor data showed a 46x productivity gap. The differentiator is now verification, not prompting.
6 min readChatGPT Work launched July 9 — an autonomous agent that connects to Slack, Google Drive, and SharePoint and delivers finished documents hours later. Three events this week prove the same principle: it works when you can define what 'done' looks like.
6 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 readFigma's CEO Dylan Field told Stratechery that AI output 'draws from the middle of the distribution' — and that teams become viscerally attached to their first concept. The iterate-and-refine skill is what separates competent AI output from distinctive work.
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 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 readThree AI failure modes got named publicly in the same week — forwarding raw AI answers without reading them, optimising for speed over quality, and letting the model think while you sign off on the result. Each one has a skill that fixes it.
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'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 readOpenAI replaced ChatGPT's default model on May 5. It's more accurate and gives 30% shorter answers. That's only useful if your prompts give it enough to work with.
4 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 readAnthropic launched native connectors to 50+ professional creative tools — Adobe Creative Cloud, Blender, Autodesk, Ableton, Canva, and more. Connecting is the easy part. Getting useful output is a constraint problem.
5 min readClaude 4.7 requires surgical specificity. GPT-5.5 demands outcome-focused framing. The same vague prompts that worked last quarter are now producing worse results on both platforms.
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 readAI drafts fast and defaults to generic. For anyone whose job is being heard, that's a real risk. Here's how to use it for speed without your writing converging on everyone else's.
5 min readAnthropic launched Claude Design this week — AI that makes prototypes, pitch decks, and marketing assets through conversation. Whether the output is useful depends entirely on the brief you write.
4 min readPrompt engineering isn't a secret vocabulary or a list of magic phrases. It's the skill of describing a task precisely enough that someone who can't ask you questions could still do it well. Here's what it actually involves.
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.