Jan Kraus

@samofsamshire @OS_Architects This is the way.

I asked AI to replace the low-rise buildings on this declining high street with @OS_Architects mock up. Imagine the increased footfall in the shops and the al fresco dining. That’s before you even consider the improved aesthetics. Visions of a better Britain. 🇬🇧

@GergelyOrosz Is happening everywhere, X, Insta, YouTube, blogs… slowly most of the content will be slop

It's now the 3rd large account on this site, just this week who used to write interesting things and about a year ago I still followed them. Now all their tweets (+threads) are very obviously AI-written, cringe to read (all the AI cliches there) and I just muted them for good.

@OS_Architects Fantastic!

We are developing some proposals for a modern mansion block with taller ceilings, more windows, better communal areas and a private garden. We’ve been researching Albert Hall Mansions and Richard Norman Shaw who is credited with making flats desirable to the aspiring Victorian middle classes @archi_tradition @Aesthetic_City @createstreets

@marclou @levelsio Before or after tax

So @levelsio was right: Stripe launched Treasury It’s like Wise but with your stripe balance. You get - bank account details - balance transfers - credit cards All with your stripe balance So you can pay friends and invoices without a bank

@UKDecline Betting shops and gambling are cancer and they should disappear.

This and William Hill 4500 job losses all in one day. This is accelerating. "William Hill and 888 owner Evoke has confirmed around 270 betting shops are being closed as it looks to offset a hit from higher taxes amid talks over a £225.3 million takeover."

@BBCNews @grok explain how this is possible. Are the companies like BP exploiting the global pricing market using it as excuse to pocket the differences

BP profits more than double as Iran war sends oil prices higher bbc.in/48qX7kU

@IroncladDev Make sense. Maybe we’ll come up with a better approach one day. Maybe it’ll just all become irrelevant at some point in the future.

@jankraus I still do PR reviews and sometimes I discover slop and point it out I give detailed reviews because I care about the software I'm shipping, but not everyone is like that

PR Review is a system that doesn't work because there are very few incentives to participate in it - It's a boring, time-consuming task that feels like a chore - Nobody commends you for reviewing code - People perceive you negatively if you are nit-picky during PR review for the most part, the reviewer is a glorified linter with some extensive knowledge of the codebase. that's it.

@bstaples How is Gitlab better?

Tired of the pain yet? Come to GitLab and take back control of your destiny. I’ll even throw in the first year free for anyone switching from GitHub who signs a new three year agreement. DM me

@ryanflorence Don’t ask “Can I…”, reframe as “How do I…”

GPT has a new phenomenon that's driving me nuts and I don't quite know how to describe it. - Ask it if I can do something - It says "no you can't [incredibly twisted restatement of what I asked but also not at all what I asked]" - It then tells me how to do the thing wonderfully - And finishes with an insulting "But you can't just [stupid thing I never actually said]" It goes something like this: "Can I form and coach a youth soccer team for my kid and play in P/D level leagues? Or do I have to be part of a full club?" Then it says: "For official competitive teams in Utah, you cannot just form a random team and enter a league. "There is a lesser-known option: UYSA allows independent teams to enter leagues if they meet requirements. [...lists some simple requirements...] "But it’s not “show up with a group of kids on game day”—it’s more like running a small club team administratively. I never said just "show up with a group of kids on game day"! It does this to me with code too. It's so weird.

@NickADobos Software engineers using code tool?

Claude is winning because rich people are providing the training data Poor meta has to train AI with the peasants of the internet Long term this will have fascinating outcomes and personality style differences as people put feedback into the system

@MissInformx4f0 @max_spero_ When I wrote some SEO slop scripts 2 years ago, I was already starting to notice this pattern

@max_spero_ I am very curious as to who was the first person to notice the not x but y ai slop form. I’m not saying for sure it was me, but I never saw anyone else criticizing the form before I noticed. So I feel good either way

This is art

@zeeg Agree. Tried on couple projects wound parade and always ended up hating it and switching to normal backend access with proper security on the backend side

RLS was a mistake and folks exposing that level of complexity to less technical users is asking for trouble. It was a mistake in Firebase. It’s a mistake in Supabase. It will be a mistake in the next product too. I personally - even knowing how to secure it - would never touch it. It’s the worst security footgun you can imagine. One small mistake and your data is available to the world.

@alxfazio FYI @levelsio

the window for experimenting with llms has basically closed now. the megacorps have fully hit escape velocity and are shipping new products and new features daily. the shift is that they’re not just shipping llms anymore, they’re using llms to build products and improve existing ones at scale. the wild west era of llms isn’t really the wild west anymore. a year ago, this could’ve been an indie dev side project, maybe even a monetizable product. it was literally so easy that the only real bottleneck was your free time. now, whatever idea you have, you should basically assume google/anthropic/oai will build some version of it within a week and wipe out most of the startup surface area around it

@GRITCULT I wish this would be true. But you’re only talking about zone 1/2

Alot of the right wing populism politics in the uk is based on fear mongering about London. Thats why they’re gonna lose again. In fact the opposite is happening Go London when the weather is good and you’ll see everyone standing out. Talking to one another. Enjoying it.

@GWHayduke97 It’s begs even broader question. What’s the purpose of that “work”?

Serious question for AI optimists who believe AI will automate all work: In a society where there are no sources of income other than UBI, and thus everyone has the same income, how would positional goods be allocated?

@GRITCULT My gut’s telling me that (but I may be wrong of course)

@jankraus Why not

Crime in the uk has fallen violence has fallen burglary has fallen Data via Office for National Statistics The UK is less violent now than it was in the 80s and 90s Dont believe everything you read on the internet UKmaxxing

@parcadei But that’s exactly what’s they’ve been accused of

ok here's my theory on why Claude Code output varies so wildly Anthropic isn't sneaking out a quantised model when nobody's looking, i think it's simpler than that so during heavy load, your request can get punted to a different server (classic distributed systems) and that server has a different batch size running (how many concurrent requests it's handling) and different batch size means the GPU switches kernel strategy so the floating-point accumulation order changes and due to floating-point non-associativity you get numerically different intermediate states these differences cascade through the forward pass so same prompt, but slightly different output + Anthropic runs Claude across multiple hardware platforms each has different precision characteristics so you're playing hardware roulette on top of the routing. the AMD director found: 5pm PST is the worst hour, 7pm PST is second worst. but why those times specifically? i think it's because these times is US evening + Asia morning collision so and you get a rampaging herd of developers signing on leading to massive load spike from global overlap and server routing goes pew pew -> cache misses spike -> you might land on different hardware -> batch size is different the kernel picks a different strategy and since the thinking budget is dynamic if you didn't set it, the reasoning depth varies too it all compounds and that's why *possibly* your Opus output is varied sometimes it's not a different model just a different path through a load-sensitive distributed system and that's why those who have more structure in the CC actually suffer less versus those who pure rawdog it and as global demand grows, unfortunately you're probably going to see more variance across all providers because it's just a nature of the game (disclaimer: this is all theory and not confirmed) below is a good read too: thinkingmachines.ai/blog/defeating…

@anorth_chen How big is your dev team?

Peter是我们的CTO，一个月前他开始实施新构建的AI-First的工作方式，结果是显而易见的，我们现在每天至少有20个PR会合并上线。产品changlog可以非常直观的体现我们团队的效率：docs.creao.ai/community-and-… 没想到他会写文章把这么干货的实践经验分享出来。 我推荐X上所有的founders好好读一下这篇文章，如果你们团队还在以AI-assisted而不是AI-first的方式去运转，很可能会在未来一年以内就逐渐淡出这个市场了。