Julian Harris

Imagine you don’t do the checks
You just hand the xref to your colleagues
You’ve created a massive cognitive liability now
You’re asking them to either trust the AI stuff or check it themselves
One of the AI project failures was the cognitive burden of asymmetric activity.

For example, I thought it would be rather clever if I used AI to create a full cross reference of one document to 4 others

The cross reference was 30 minutes of work

But it was a whole day to do the checks to make sure it was all real

Fix formatting issues, misunderstandings etc

It requires quite fundamentally different skills on top of your normal IC skills.

And there are some really big risks too.

Sometimes it’s like using a bazooka to SWAT a fly

I think it’s just a significant behaviour change

It’s aa significant as it is from someone to move from being an individual contributor to being a manager.

6 Hard Truths Behind MIT's Finding That 95% of AI Pilots Fail Explore 6 hard truths behind why 95% of AI pilots fail and uncover strategies to ensure your AI projects succeed.

Partly. Also readiness, and focus among other things.
www.cloudfactory.com/blog/6-hard-...

Thanks! Probably the use case I shared — lead qualification.

Bluesky needs growth, not polls.

hey jp you called it

My measure of AI coding agent performance is inverse to the number of swear words I dole out to them in a chat.

Interestingly, with Sonnet 4 Max on Cursor my profanities have been surprisingly sparse. But then also it cost me $80 from one day of usage 😱

AI coding agents accelerate code creation. It doesn’t accelerate organisations.

Same

How do you get started with agents in your business? I answer this with a super in-depth deep dive, step by step.

makingaiagents.substack.com/p/the-best-n...

The cover of my upcoming book on Vibe Coding. What do you think?

Ah I have a sibling version of that.

Interesting! Doesn’t seem to touch on schema migrations though (I could have missed it?)

There’s at least one company now moving from literal retrofits to entirely new cars inspired by yesteryear designs.

I watched a 15 minute YouTube video on this so I’m something of an expert on the topic. Biggest issue is retrofitting batteries: you can’t really sneak them into crevasses and they’re really heavy so it’s better to have them low, flat and centred.

This is all I can see on mobile

Chrome has a neat mobile device test mode.

Is there any decent ice cream in the US? Seems like the US palate has a stronger preference for sweetness than I’ve experienced elsewhere.

Nice. Could you add mobile support? It’s very hard to use on mobile currently.

Why can’t the ground truth be built in?

Seems like the message is “you cannot fully automate evals yet, human in the loop is still needed” — correct?

Mortgage free house?

10 like tickets please 👍👍👍👍👍👍👍👍👍

Funny guy

Yes, if we look at the conventional axes of faster cheaper better, then what does that look like? And is there an equilibrium point where you can’t really get faster cheaper better. What’s the CD audio moment? What is the retina display moment?

Yeah strange slightly breathless feeling isn’t it when the whole solution is implemented in seconds correctly despite having spent maybe hours coming up with the prompt.

source?

So much fun. Are we witnessing a renaissance in the game industry? I saw a post that suggested this early raps innovation is how it was in the good old days of gaming.

No Mac version is that because Steam?