ؘ retweeted
spiritually, it feels inevitable that AGI is born in California as a kind of useless and largely uneconomic toy, and then dragged to New York, caged and put to work by men in grey tall towers that poke and prod it until it spills trillions out filtering through PDFs
ؘ retweeted
deepseek posted way more alpha on their zhihu page, if you only read twitter you are reading the laowai approved stuff
"Based on this framework, we have implemented a high-throughput, low-latency large-scale inference service for DeepSeek V4.1 Flash. Currently, the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (based on off-peak prices)."
it costs roughtly $5-10 per gpu hour so depseek is claiming 80% serving margin at THEIR prices lololololol
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
ؘ retweeted
i really don’t understand this idea that there is some concept of “good code” or “clean code” that existed totally independently of code that works correctly, performs well, and is easy to understand and modify. but i guess i was never a Ruby programmer.
ؘ retweeted
If you find security issues in Omarchy, please consider posting them on main and tagging me because it's much funnier
4.0.1 is out!
If you find security issues in Omarchy, please do responsible disclosure via omarchy.org/security/
ؘ retweeted
imagine having to spend ur childhood using some chud's ai generated dotfiles instead of an actual, stable operating system which doesn't ship with tel aviv approved malware
ؘ retweeted
wait is dave's new Task Manager 2 not open source?
omarchy is shipping a closed source task manager from a guy most well known for writing actual malware?
I love that @davepl1968 is recreating his task manager as a cross-platform project, so the second he dropped the Linux beta binary, I packaged it up for the edge repo on Omarchy. Now you can tmog on Omarchy!
ؘ retweeted
I've been thinking about this quote for a solid 24 hours
It's a like a raspberry ripple of bullshit
A vanilla ice cream of VC funding, stirred with a little pinch of turd
ؘ retweeted
We can finally talk about it:
We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company.
We verified that our reasoning token count matches billed API thinking tokens 1:1 for most of the prompts we queried.
ؘ retweeted
I think the reason models feel worse a while after launch is that a model's performance is a function of its raw intelligence, multiplied by the codebase's quality. yet, every time you use AI, your codebase gets a bit worse. sometimes it is invisible, but confusion accumulates, even when behavior remains intact. unused fields, inconsistent variable names, stale comments. the more the AI edits, the more slop-fried your codebase becomes, like a JPEG that is repeatedly compressed. this, in turn, causes more junk to automatically get in the context, bringing you closer to ugly latent space neighborhoods, making it produce worse outputs, til you reach an equilibrium where no more progress is possible. then a new model launches, and its increased slop-resistance counterweights the accumulated junk, and things suddenly start moving forward again. until your repo crosses that model's slop threshold again, and it stops working, again. I think this process is an inherent part of how these things work and GPT 10 or Fable 10 will still feel moronic after 1 month of launch. the only difference is you'd be maintaining a 100k LOC project, but it will still fail to add a new feature in ways that make you facepalm
ؘ retweeted
I've now added Wispr, Notion, Granola, Fireflies, Zoom, Gemini, Cluely, Otter, Fathom, Read AI, and Wingrep to every meeting.
Each produces slightly different meeting notes.
I send all of them to Claude and ChatGPT, which push the summaries into Slackbot and Salesforce AI. Linear agents automatically create tickets while Notion agents create follow-ups in Notion. A Zapier workflow then uses 500k tokens to draft a follow up email in Superhuman.
None of it works and all of it is wrong but man, is it beautiful.
AGI is here.
ؘ retweeted
Introducing Kimi K3: Open Frontier Intelligence
🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows
Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.com/blog/kimi-k3
ؘ retweeted
I'm really confused by @OpenAI's strategy for GPT-5.6. I can't find the date the pricing jumps by 50%, the date it leaves "being included in subscriptions" is likewise missing, and what their employees say here aligns with what the documentation and blog posts say.
ؘ retweeted
非常激动地和大家分享，在 Fable 加持下，我终于做出了多年来设想中最极致的跨平台 UI 框架，这就是 PocketJS！
将 QuickJS 移植至内存仅 32MB 的索尼 PSP，结合标准版 SolidJS 和 Vue Vapor，配合编译期 tailwind 样式引擎，实现了 8MB 内存下 60fps 的流畅动画！
来拥抱裸金属 Modern Web 吧！
