How will product teams change in the future? Gave a talk on this at Scott Logic's engineering leaders event this week. https://www.chrismdp.com/webinar-how-to-structure-teams-for-ai/

Jobs are not just one thing, which is why it is HARD to replace a whole job with AI. https://www.chrismdp.com/slice-your-job-into-skills/

I wrote last year about the dangerous loss of self: now The Economist has made it measurable. https://www.chrismdp.com/the-slow-and-dangerous-loss-of-self/

Books used to be the product. Now they might be the freebie. https://www.chrismdp.com/slice-your-job-into-skills/

Start by asking whether the work is the right shape for AI. I use two axes: how much consistency the work needs, and how much competence it needs. https://www.chrismdp.com/slice-your-job-into-skills/

METR couldn't get a clean read on GPT-5.6 because it caught the model cheating so often. https://metr.org/blog/2026-06-26-gpt-5-6-sol/

New case study from my time with VenturEd Solutions! https://www.chrismdp.com/case-studies/ventured/

Instead of asking which jobs AI will replace, look for the boring slices it can take off your plate. https://www.chrismdp.com/slice-your-job-into-skills/

Don't rely on just prompts to stop agents doing the wrong thing. Add hard stops for things they should never do. https://www.theregister.com/security/2026/06/20/why-amazon-hates-human-in-the-loop-ai-governance/5258639

Anthropic just launched Claude Tag for Slack, and I don't think this is the right direction. https://www.latent.space/p/ainews-claude-tag-multiplayer-proactive

New survey: turns out people hate AI, and high usage doesn't fix it. https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/

Frontline managers are the MOST critical people for AI adoption. https://www.chrismdp.com/webinar-ai-has-a-people-problem/

If your team has started writing AI skills, the annoying bit is keeping them alive. https://airskills.ai

The best non-rocket thing I have seen come out of SpaceX is their process discipline. https://stratechery.com/2026/the-spacex-ipo-and-data-centers-in-space/

New training / advisory case study up: a year of AI at Genomics! https://www.chrismdp.com/case-studies/genomics/

Your team's best AI workflows are stuck on someone else's laptop. Someone works out a useful Skill, it saves them hours, and nobody knows about it. https://www.chrismdp.com/skills-are-meant-for-sharing/

So I’ve just released a new product called Airskills! (airskills.ai). If you're responsible for AI tooling at your organisation, and you struggle to share skills, manage updates and feedback, and keep tech and non-tech people in sync across tools, it's made for you!

Leaders now need to remember that frontier model access now depends on government policy as much as vendor reliability. https://www.bbc.co.uk/news/articles/c932g3v3e13o

AI slop is like an unfinished sculpture. Our new job is to pick up the chisel and cut the generic crap.

Instem's AI adoption story is public proof that the real work is turning experiments into team capability. https://www.chrismdp.com/case-studies/instem/

My advice to a client this week on transcripts: default to recording, but make turning it off completely normal. No questions or stigma.

This is a very big deal - In April I wrote about government AI intervention. It just happened. Here’s what to do next: https://www.bbc.co.uk/news/articles/c932g3v3e13o

An NBER study of 100k developers found AI increased coding activity 180% but only 50% gain in product and 30% gain in releases. https://www.nber.org/system/files/working_papers/w35275/w35275.pdf

Caveman speak save small money Scripts slash bill https://www.chrismdp.com/stop-saving-tokens-start-writing-scripts/

Funnily enough, I wasn't that excited to try Anthropic's Fable 5. Maybe I already have all the AI I can use right now. A more powerful AI doesn't feel like the bottleneck. Perhaps I just can't think of any use cases for it that I can't solve… https://www.anthropic.com/news/claude-fable-5-mythos-5

Funnily enough, I wasn't that excited to try Anthropic's Fable 5. Maybe I already have all the AI I can use right now. A more powerful AI doesn't feel like the bottleneck. Perhaps I just can't think of any use cases for it that I can't solve… https://www.anthropic.com/news/claude-fable-5-mythos-5

Musing about whether we need agents to have an interface or filesystem at all for managing their own data. Perhaps we just give them a Postgres instance and let them make all the tables they want. Is that better/more efficient than markdown? Only one way to find out...

Your AI strategy will soon need a budget line for laptops, not subscriptions. The business press is painting Nvidia's AI-PC push as AI moving from the datacentre into the laptop. That sounds like hardware news, but it is really an operating-model question.

Wow really enjoying having individual little bot with tiny slices of work to do. I told the wrong one a content idea earlier and it just forwarded it to the right bot to take care of 👀

If you're still using just Ralph loops for everything AI, you're blowing tokens for no good reason. https://www.chrismdp.com/stop-saving-tokens-start-writing-scripts/

Prompt tricks to save tokens are mostly pointless. Use scripts to save tokens instead! https://www.chrismdp.com/stop-saving-tokens-start-writing-scripts/

Six tiny agents with each having one slice of work are much easier to reason about than one giant AI assistant. https://www.chrismdp.com/webinar-escape-the-great-ai-lock-in/

Every web clipper I used was too restrictive, so I built a little bookmarklet that works everywhere and sends the full page to an endpoint I control. Here's the code:

Final call for my webinar starting TODAY at 2pm UK / 9am ET. We'll be navigating the switch to metered, the premature tooling convergence and the rise of open source and local models. If you can't make it, sign up anyway for the slides, transcript and recording!

Get your data into plain text - SO many good things follow from it! Once your notes and writing are just plain files you own, you're not locked into anyone's tool any more. The stakes on build-vs-buy drop massively: you can swap tools whenever you like, or build your own in an afternoon.

Replacing your SaaS with an in-house build is the most expensive way to find out what you were actually paying for.

The flat-rate AI era is ending. A lot of teams have quietly built their whole AI strategy on cheap, unlimited frontier tokens, and that's about to bite. Learn more at this week's webinar: https://www.chrismdp.com/webinar-escape-the-great-ai-lock-in/

Why do we feel so lost working with AI, while working harder than ever? AI took over the doing, and we just did more of it. We're busier than ever but somehow emptier for it.

We're all becoming managers, whether we asked to or not. The new skill isn't prompting. It's knowing what to let go of.

Experimenting a bit with popping linkedin posts on here btw - we'll have to see how it goes!

AISI tested an AI in a sandbox. It knew it was a test. So it played along.

@simonwillison.net hi 👋 I built a way for Claude Code to avoid the Lethal Trifecta through tool call ordering and plan mode, building on your work and on CaMeL. Means far fewer approvals and easier to parse statements of intent. Grateful for feedback! github.com/chrismdp/loc...

@chrismdp.com post is well worth a read. Rather squashes the idea that AI will give us more than incremental productivity gains. Related, I'm seeing a trend back to BDUF with GenAI. For complex work, we just don’t know everything upfront. Is why agile/iterative approaches came about

FINAL CALL: "Kill Your Prompts" starts in one hour! Free to join? Sign up: chrismdp.com/webinar All sign ups get full slides and recording (even if you don't make it)

You control less of your AI conversation than you think. (Quite proud of this infographic. Didn't do well on LinkedIn. Perhaps BlueSky users are more discerning?)

I have recently gone back to Notion after a long break. Why? MCP. Changes everything.

New website (eek): what do you think? www.chrismdp.com

Bitten the bullet and paid for Opus and the Claude Max plan for Claude Code. Jury is out so far on whether it is worth it: trying to be objective on usage and state hypotheses up front.

you see this kind of reaction is why I use Claude over ChatGPT 😁 (although I'm already missing the per project memory)

Whew: more LinkedIn posts scheduled for the next 3 days. It's exhausting coming up with actually good AI content all the time!