Originally UK, now Montreal. Stuck in 1998.

Montreal
Joined February 2014
The synth is out, you can try it at crazyphase.com/czp-1/ and there is more info at the blog.
1
208
Notably the developers I know now have all given up on believing AI won’t ultimately be better than them, if it isn’t already. Software is no longer valuable by itself. If you want to run a software business you must have critical features enabled by data you keep secret.
2
2
36
Web 2.0 redux, basically. The community is the product, you get more valuable with more users etc.
13
atomirex retweeted
say what you want about @elonmusk, but I can guarantee you ain't building: - reusable rockets - self driving cars - flying cars - frontier AI - satellites - tunnels - robots - solar - X honoured to be living in the same timeline as him.
1
2
9
411
Aside from anything else, the most boring application of this would be radically improving safety by providing a way for cars to get out of the way and do a controlled stop in almost any situation. It could end up being mandatory.
Turns out the future does have flying cars
32
I have become a classic product manager, pretending a feature is no longer happening so the minions focus on the part they need to do.
3
37
Found my lovely cat Groovy’s gashapon in Koenji tonight!
1
7
21
764
To give a concrete example of latest Sol madness I saw it add in CPP if(variable <= CONSTANT_1 && variable <= CONSTANT_2) etc. ignoring that CONSTANT_2 is guaranteed to be larger than CONSTANT_1.
1
30
This latest round of second tier models (Opus, Sol etc.) are deceptively dumb. The joy of Fable and Astra is they sometimes act like they can reason beyond the words, but even while Opus and Sol are astounding at structured regurgitation their reasoning is just non existent.
1
2
53
The giveaway for this is the inability they have for functioning in situations without closed feedback loops involving very strict pre-existing requirements. In those they are amazing, but outside it becomes a total mess.
1
2
13
Replying to @atomirex
I have heard a tale first hand from an engineer who claimed an agent refactored something and simplified the codebase this one time. And we all laughed at him.
1
1
17
I've seen these coding agents do many wondrous things, but not once have I seen one successfully simplify anything at all.
2
1
35
The whole point of AI is to enable maximal creativity. Aping historical signals of production value misses the point because it just looks like low effort AI now, even if it's not.
1
25
A framework is an intelligence cache. You could waste tokens reinventing the ORM, cache, queues, auth, routing, etc, or you could just pick a world-class framework.
I don’t know if I agree with this take. Good frameworks are intelligence caches. The production of verifiable high-quality code has a cost to it (insert for now caveat). Frameworks can lower this cost as long as they don’t get in the way of agents.
78
86
35
1,066
48,951
Figma acting like a synthesizer company in 1984 that only wants their MIDI implementation to work with those from companies they preapprove, then acting surprised when the ecosystem proves to be bigger than any single product.
Replying to @GayaniFigma
Hey @GayaniFigma , I think in spirit, when I created MCP, I envisioned an open ecosystem. That to me feels core. Seeing restrictions like this is sad and I hope Figma can get to a point where it’s more open or at least make the process of getting into the allowlist very easy.
2
105
The whole "agents can port code now" thing is cool. Let's see them maintain all the resulting ports in sync afterwards.
27
Don't entirely agree with Andreas in his ongoing threads, but he has lately been thinking a lot further ahead than most on open source processes, security etc. and what all that will look like. The part being glossed over here is how users will know which projects can be trusted.
Zoomers who grew up in the GitHub era keep telling me that closing PRs means projects are no longer open source. Obviously, all the requirements for open source still apply. Public code intake was never part of the definition; it was just a popular contribution model for a time.
1
3
320
Quoting from the report; 'Robots and LLMs together expose all but one-fifth of employment.'
3
3
52
5,387
The big mistake with agents today is they increase communication overhead and noise. Every additional agent makes it exponentially worse. Co-ordination via chatroom simply doesn't scale up, whether with humans or machines, let alone both.
1
48
Local man confuses labor for outputs
really don't see the appeal to "vibe coding" a game... you don't learn how anything works or get hands on experience building something, you get a sub par result & can't take credit for it. why is it always about skipping the important steps just to get validation from strangers
1
3
131