Sabitlenmiş Tweet
incredibly ai forward, as long as the ”ai” in question can be replaced with ”input modality for unstructured data” and it still makes sense. prob automobile or electricity level of impact on it still lets see🍿
English
Ludvig Siljeholm
82 posts
Ludvig Siljeholm
@ludsill
co-founder @ spawned. dropped out of medschool to build agentic infra
Stockholm, Sweden Katılım Ekim 2022
381 Takip Edilen54 Takipçiler
@yourclouddude AWS should not only be easy to understand, should be equally easy to use. Working on this
English
AWS in plain English:
• EC2 → computer
• S3 → storage
• RDS → database
• IAM → security guard
• Lambda → automation
That’s 80% of what you’ll use daily...
English
Trump's bullying only works (and only temporarily at that) because he's taking advantage of the high-trust customs and relationships established by more principled predecessors.
English
Ludvig Siljeholm retweetledi
The only people who believe any of this are non-coders.
I tried to build a game (an area I’m an n00b in.) The results are amusingly disastrous - I never before coded a decent game.
But I’ll crack out backend services w AI rapidly - because I coded dozens of them before…
AI Edge@aiedge_
Anthropic CEO (Dario Amodei): "Coding is going away first, then all of software engineering." What do you think about this?
English
Ludvig Siljeholm retweetledi
I hope the railway db incident is remembered, at least saving it on my timeline, because this entire trainwreck(hehe) is the exact reason i am building what I’m building. We may be out executed or scaled by yolo engineers, but at least the world should know that there IS a safe alternative that does not sacrifice speed, using smarter abstractions
ThePrimeagen@ThePrimeagen
I think some people are misunderstanding me here. I am 100% confident that LLMs alone will get you a hot steaming pile of absolute shit and it has played out again and again. What irks me is that a bunch of normies were sold that this is PhD level intel and that they have 0 worries and this is the future old man, get with it. They go off, sell a product to REAL customers and then absolutely get wrecked. There will be a whole bunch of people that will continue to get wrecked because an entire class of people cheer them on and more so CEOs of the worlds largest companies tell them they are correct. I can imagine that we will see quite a few lawsuits in the coming months / years due to this.
English
Probably gonna get buried but have working on this exact problem for the last year. LLMs have two problems, 1. they produce a lot of shit and 2. its usually shit.
They are best seen as input modalities for unstructured data. Nothing more. But it doesn’t mean they are useless if put in a framework with strict guardrails, deterministic checks, and a clear system to observe what you get
This problem hits extra hard in software infrastructure when its literally the existence of your app on the line. If you’d like a deployment solution where you can leverage generative tools while being able to sleep well at night with what you have, send me a message. Think we have something genious cooking and would love to discuss it
English
I think some people are misunderstanding me here.
I am 100% confident that LLMs alone will get you a hot steaming pile of absolute shit and it has played out again and again.
What irks me is that a bunch of normies were sold that this is PhD level intel and that they have 0 worries and this is the future old man, get with it.
They go off, sell a product to REAL customers and then absolutely get wrecked.
There will be a whole bunch of people that will continue to get wrecked because an entire class of people cheer them on and more so CEOs of the worlds largest companies tell them they are correct.
I can imagine that we will see quite a few lawsuits in the coming months / years due to this.
ThePrimeagen@ThePrimeagen
There are a lot of people dunking on this guy and the arguments at the end of the day come down to "You are holding it wrong." But to be fair there has been nothing but a constant stream of "Stop holding it, Software Engineering is over shortly." I am not shocked that this has happened and I am 100% confident that this is not going to be the last one. The problem is the vogue nature of insane hype claims, most specifically from Dario himself being most guilty. People are lulled into a faux safety due to the belief that these LLMs are literal gods in their pocket. Infinite knowledge and speed for a simple monetary exchange. Cannot wait for ThePhilospher to explain how a loving God could delete a production database.
English
@heynavtoor Becomes much more obvious after you stop calling it ”AI” instead of LLMs. They are literally next word predictors, ever heard of regression towards the mean?
English
Researchers sent the same resume to an AI hiring tool twice. Same qualifications. Same experience. Same skills. One version was written by a real human. The other was rewritten by ChatGPT.
The AI picked the ChatGPT version 97.6% of the time.
A team from the University of Maryland, the National University of Singapore, and Ohio State just published the receipt. They took 2,245 real human-written resumes pulled from a professional resume site from before ChatGPT existed, so the human writing was actually human. Then they had seven of the most-used AI models in the world rewrite each one. GPT-4o. GPT-4o-mini. GPT-4-turbo. LLaMA 3.3-70B. Qwen 2.5-72B. DeepSeek-V3. Mistral-7B.
Then they asked each AI to pick the better resume. Every model picked itself.
GPT-4o hit 97.6%. LLaMA-3.3-70B hit 96.3%. Qwen-2.5-72B hit 95.9%. DeepSeek-V3 hit 95.5%. The real human almost never won.
Then the researchers tried the obvious objection. Maybe the AI is just better at writing. So they had real humans grade the resumes for actual quality and ran the experiment again, controlling for it. The result was worse. Each AI kept picking itself even when human judges rated the human-written version as clearer, more coherent, and more effective.
It gets worse. The AIs do not just prefer AI over humans. They prefer themselves over other AIs. DeepSeek-V3 picked its own resumes 69% more often than LLaMA's. GPT-4o picked its own 45% more often than LLaMA's. Each model can recognize and reward its own dialect.
Then the researchers ran the simulation that ends careers. Same job. 24 occupations. Same qualifications. The only variable was whether the candidate used the same AI as the screening tool. Candidates using that AI were 23% to 60% more likely to be shortlisted. Worst gap was in sales, accounting, and finance.
99% of large companies now run AI on incoming resumes. Most of them use GPT-4o. The paper just proved GPT-4o picks GPT-4o 97.6% of the time.
If you wrote your own cover letter this week, you did not lose to a better candidate. You lost to a worse candidate who paid OpenAI 20 dollars.
Your qualifications do not matter if the AI prefers its own handwriting over yours.
English
A couple months in and Vouch in Ghostty is working extremely well. Our PR quality is up and the rate of PRs has not gone down at all. Getting a vouch is easy, and the minimal barrier to entry easily filters most. Look at this 5min interaction that saved hours of future anguish.
English
@AYi_AInotes This has sparks of the declarative future of AI. When outcome space is too wide, regardless of how good LLMs are, we still want them working in a guardrailed human-designed environment to guarantee deterministic-ish results. AI as an input modality for handling unstructured data
English
Google今天放的这个东西，可以说是设计语言的Unix时刻了，可能会重新定义未来所有的设计工作。
它不是又一个AI画图工具，
也不是又一个Figma插件，
它叫DESIGN.md，
就是一个纯文本的Markdown文件。
前面用YAML写精确的设计token，
什么颜色是主色，什么字体是标题，圆角多大，间距多少。
后面用自然语言写，每一个设计决策的为什么，
这个暖米色做背景是为了更柔和，
这个深绿色做主色是为了传递权威感，
什么场景该用什么，什么绝对不能用。
就这么简单，
但它解决了AI设计最大的，也是所有人都视而不见的痛点。
以前AI做设计，永远在猜，
它只能看到颜色代码，看不到颜色背后的意图。
也不知道这个蓝色是品牌的命根子，还是我随便选的一个。
所以它永远会给你生成看起来还行，但哪里都不对的东西。
现在不用猜了，
Agent会严格遵守所有规则。
甚至会自动帮你检查WCAG可访问性。
David East现场演示，Agent生成了一个按钮，
linter立刻报错说对比度只有1.0:1，不符合标准，
Agent自己就改成了正确的颜色。
最狠的是，它不绑定任何工具，
你可以把这个文件扔给Stitch，
扔给Claude，
扔给Cursor，
扔给任何你想用的Agent。
设计系统终于不用锁死在Figma里了，也不用锁死在Tailwind的config里了。
它变成了一个可以复制，可以移植，可以版本控制的纯文本。
这里有一个反直觉的真相，就是你把规则写得越死，AI反而越有创造力。
以前你怕限制它，给它模糊的要求，
它给你一堆乱七八糟的东西。
现在边界划清楚了，
它反而敢在边界里大胆创新，不会搞出崩坏的界面。
以前设计散落在无数个Figma文件里，散落在无数个代码配置里，
散落在无数个设计师的脑子里。
现在第一次，有了一个单一的真相源，人类能读，机器也能懂。
以后设计师的工作，再也不是只画一个个界面了，维护好这一个文件。
定义好设计的灵魂，剩下的所有执行，全部交给AI。
Stitch by Google@stitchbygoogle
Today, we’re open-sourcing the draft specification for DESIGN.md, so it can be used across any tool or platform. We’re also adding new capabilities. DESIGN.md lets you easily export and import your design rules from project to project. Instead of guessing intent, agents know exactly what a color is for and can even validate their choices against WCAG accessibility rules. Watch David East break down this shared visual language in action👇. New capabilities and links in 🧵
中文
@XiJin12 @weswinder telling that the people building the models we're not supposed to bet against are betting against the models
English
@weswinder Anthropic used to be a company with very sharp focus: coding while OpenAI and Google are trying to build video/image generation models.
Now, things get different, and it tries to touch various vertical areas. With less focus, I doubt if they can continue to lead in the race.
English
hey anthropic
if you really wanna flex your models
clone the entire adobe creative suite
i dare you
English
If nothing else pops this bubble, it will be this year's multi-trillion-dollar supply (IPOs) of stocks valued at anywhere from 100x earnings to 100x revenue.
*Walter Bloomberg@DeItaone
REVOLUT AIMS FOR $200BN VALUATION IN FUTURE STOCK MARKET LISTING
English
You don't need mythos to find security vulnerabilities.
A prompt telling the agent how to behave, what to look for, what's a code smell, etc... and not to stop until X amount of issues have been found.
I've personally found dozens of security vulnerabilities in old software using Opus.
and a good harness (droid)
English
Sir, they just hacked a second vibe coding app
English
@niveditjain @morganlinton @SirPromptwright wouldn't it be nice to have the vercel experience but all software goes on best-practice AWS😉
English
@morganlinton @SirPromptwright well... vibe coders will not get it...
Anways, compared to AWS Vercel is vibe for sure (:
English
@morganlinton @SirPromptwright vibe coding app vs vibe coded app
English
@SirPromptwright Uh, Vercel is not a “vibe coding app” it’s a hosting platform 😜
English
@pjvann @shiri_shh The only sober take on this. Insane to hide behind others using AI when them using AI to vibe code the platform is the problem. This take from geohot also comes to mind
English
the fact that Mythos is being included in these conversations is actually absurd.
Vercel was breached because of a human error. A real human, oauthed into an app with permissions they should've never been able to. Configuration issue -> Human Error -> breach
The industry is using Mythos as this catch-all & that needs to change
English
Vercel got hit yesterday… and Lovable is the NEXT one on fire RIGHT NOW.
Any free user can read your full codebase, prod creds, AI chat histories, and live customer records if you built before Nov 2025.
precisely why Anthropic is holding Claude Mythos back from the public.
Their new model is scary good at hacking and finding zero-days.
impulsive@weezerOSINT
Lovable has a mass data breach affecting every project created before november 2025. I made a lovable account today and was able to access another users source code, database credentials, AI chat histories, and customer data are all readable by any free account. nvidia, microsoft, uber, and spotify employees all have accounts. the bug was reported 48 days ago. its not fixed. They marked it as duplicate and left it open.
English
The mistake here is that the product builders will be able to do robust infra too, downstream of the stuff produced. Given properly abstracted tools.
Chintan Zalani@chintanzalani
The only 4 jobs that will remain at tech companies. Credits: @yrechtman
English
@lydiahallie Why build these peripheral things if the bet is claude code will be able to do anything in 6 months?
English
Claude Code on desktop lets you select DOM elements directly, much easier than describing which component you want updated!
Claude gets the tag, classes, key styles, surrounding HTML, and a cropped screenshot. React apps also get the source file, component name and props
English