This could be used for a truly eye-opening art installation: a screen that as you walk by it, tells you when you were last there..

Even wilder would be to buy data on you in real time and display that.

Great to see people thinking about this. But it feels like a step on the road to something simpler.

For example, web accessibility has potential as a starting point for making actions automatable, with the advantage that the automatable things are visible to humans, so are less likely to drift / break over time.

Any work happening in that space?

If anyone is looking for ideas for these projects - it’d be great to be able to run macos applications on linux…

Someone could have a swarm of agents build “wine for macos apps”.

> The goal of an LLM is to continue the conversation.

It’s even simpler. The goal of an LLM is to generate the next token.

That’s reductive but worth considering. An LLM doesn’t have inherent goals and you aren’t privy to how it was post-trained or what on, so you can’t assume it’ll behave in any particular way.

What an amazing article. I feel like my brain has been cleansed of a bunch of old assumptions.

Thank you to the author.

No one else has done it and code is easier than ever to create. This tool needs to be built by the person closest to the problem.

Ask your agent for ways to do this using code, not more AI.

It might propose - and build! - an embeddings based system and scraper for your issues & PRs. Using that will burn zero tokens and you can iterate on it as you think of improvements.

AI won’t kill subscriptions but it will drive prices down. Viable competitors will emerge and incumbency will become a less viable long term play.

Non-experts will build progressively more interesting things as model capability goes up - but experts will always be able to produce yet more complex things than that.

So user expectations will keep rising.

The cost of software development will go down for software at and below some level of complexity, but up for anything above that level. The level where that tips will keep changing, just as it always has. To understand this, consider that tipping point across today’s and yesterday’s “hand-typed software”: it has been moving the whole time, thanks to frameworks and libraries and languages improving.

The world of software will remain much as it is today - a mix of free and non-free stuff. But on average and in the large it’ll be of way higher quality, and bring corresponding improvements in your quality of life.

Thanks, but asking AI for anything verbatim is risky.

You’d be better served asking the bot to scour the internet for a link to a source and reading that yourself.

They are in some other democracies. We will see how it all plays out.

If after 7 or 8 weeks you can’t change the software, it’s wise to start putting tests in to document retroactively how things work / worked and why.

The more the test suite grows, the more you & future agents will be able to consult it to understand how things should work - but also, why.

Imagine a test case that covers some non-compliant API response from a third party. The commit that’s tied to, the date the test was added, all that becomes metadata.. and the fact it’s executable means your agent can’t undo that fix without something very visible in the PR.