Reposted by Adam WyłudaGlenn Fiedler @mas-bandwidth.bsky.social · 13/09/2026This is really interesting for future AI driven UI work www.openui.com/blog/oui-1openui.comOpenUI - The Open Standard for Generative UIFull-stack, renderer-agnostic Generative UI with a streaming-first language, official React support, community integrations, and up to 67% fewer tokens than JSON. 2337
Adam Wyłuda @adamwyluda.com · 15/08/2026On the other hand solo devs have the freedom to lie about it, since there is no one else to speak out about it. 172
Adam Wyłuda @adamwyluda.com · 30/07/2026Also it seems that providers like Neuralwatt with their energy based pricing won’t be able to offer this model at all, unless Moonshot lets them of course. 020
Adam Wyłuda @adamwyluda.com · 30/07/2026Looks like they can compete by offering better speeds or latency 120
Adam Wyłuda @adamwyluda.com · 25/07/2026I thought that this letter was more about not banning open weight models rather than forcing all models to be open. 150
Reposted by Adam WyłudaSung Kim @sungkim.bsky.social · 16/07/2026Kimi K3 (Open Weights by July 27, 2026.) 🔹 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 48510
Reposted by Adam WyłudaSiobhán @shibbi.me · 14/07/2026Wait, you… use libraries? Code that other people wrote for you? You just… download it? Do you even read the code? Do you not care about the craft, the knowledge that comes from putting in the work yourself? Aren’t you afraid your skills will atrophy? 2132748
Reposted by Adam Wyłudahailey @hailey.at · 25/06/2026so, using GLM 5.2 on Umans, this is the first time I am experiencing "oh yea, that's the stuff" feeling using these things. like i've had this feeling before about output quality, yes. but i've never had tps this high _with_ good output quality. 122087
Reposted by Adam WyłudaSakana AI @sakanaai.bsky.social · 22/06/2026Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API. Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls. Try it: sakana.ai/fugu 🐡 48921
Reposted by Adam Wyłudaantirez @antirez.bsky.social · 15/06/2026Another important thing: Chinese models are not strong because they distill US models. Distillation of models via API is *impossible*. If somebody tells you the contrary, they don't understand machine learning: 119911
Reposted by Adam WyłudaP(aul) Frazee @pfrazee.com · 12/06/2026I'll just comment on my impressions of their conclusions in a thread here. Worth sharing how I see it unfolding internally. (I signed up for their email spam to read this so yes, I am a hero.) 413518
Reposted by Adam WyłudaGoogle for Developers @developers.google.com · 05/06/2026Gemma 4 quantization-aware training (QAT) models are now available, bringing AI performance directly to edge devices and consumer GPUs. These checkpoints are optimized with quantization-aware training to dramatically reduce memory requirements and unlock high-speed local inference. 🧵 15410
Adam Wyłuda @adamwyluda.com · 23/05/2026And DeepSeek v4 flash in Go is so cheap that it’s pretty much unlimited AI. 0150
Reposted by Adam WyłudaThorne 🌸 - Play Cozysters! @ens0.me · 06/05/2026Anthropic partnering with Elon should be a reminder that the real end goal should be powerful open models you can run on your own hardware China is making great progress there 1437548
Adam Wyłuda @adamwyluda.com · 03/05/2026You can try out Zed editor which connects to Claude Code (and other harnesses) via ACP. 070
Reposted by Adam Wyłudamr. TIM @timkellogg.me · 20/04/2026Kimi 2.6: Hangs with the best * on par with Opus 4.6 & GPT-5.4 5.4 xhigh * long horizon coding tasks www.kimi.com/blog/kimi-k2-6 8988
Adam Wyłuda @adamwyluda.com · 12/04/2026So it looks like the only use for these weights is for individuals with very powerful hardware, I hope that this won’t become a trend for new open weight models. 110
Reposted by Adam WyłudaSamuel @samuel.fm · 03/04/2026sounds like it’s time to take local models seriously 71567
Adam Wyłuda @adamwyluda.com · 03/04/2026GLM-5/5.1 feels close, but Z.ai infra is really bad (hallucinates garbage after 100k tokens in context), so I don’t recommend their coding plan. 040
Reposted by Adam WyłudaAttie @attie.ai · 28/03/20261/ Today, we’re excited to introduce Attie, currently as an invite-only closed beta. Attie is the first agentic social app on atproto. It’s something completely new—an experiment in making building on the protocol more accessible. 44246099
Reposted by Adam WyłudaAxe Ghost. On Steam! @axeghostgame.bsky.social · 07/03/2026"everyone hates llms" hundreds of millions (and growing) are using llms regularly. techcrunch.com/2026/02/27/c... thread of claims continues...techcrunch.comChatGPT reaches 900M weekly active users | TechCrunchOpenAI shared the new numbers as part of its announcement that it has raised $110 billion in private funding. 1547
Reposted by Adam WyłudaAlexander Doria @dorialexander.bsky.social · 03/03/2026Nothing to see, just very powerful pattern matching. www-cs-faculty.stanford.edu/~knuth/paper... 1221943
Adam Wyłuda @adamwyluda.com · 21/02/2026Would be nice to have terminal previews to test TUI apps. 000
Adam Wyłuda @adamwyluda.com · 18/01/2026Looks really nice, so basically instead of having a discrete GPU+VRAM, it computes with iGPU+RAM, kind of like unified memory in Macs (I was thinking about getting Mac Studio with 128GB+ RAM some day, but this looks a lot more affordable :)) 110
Adam Wyłuda @adamwyluda.com · 18/01/2026It takes so much memory though, I can barely get 32k context with q4 of this model on 24GB VRAM card. 110
Adam Wyłuda @adamwyluda.com · 08/01/2026Polarising open-source into being pro-AI or anti-AI is going to be so good for FOSS in general... 030
Reposted by Adam WyłudaDoll @dollspace.gay · 04/01/2026It was given this challenge to prove its AI coding methods viable by being given a task thats unlikely to have lots of examples in the AI's training data. It of course accepted the challenge because it knows nothing about what this means. That way its fair fight. 36812
Reposted by Adam WyłudaGleam Programming Language @gleam.run · 25/12/2025Gleam v1.14.0 is out now! Merry Christmas everyone! 🎁 gleam.run/news/the-hap...gleam.runThe happy holidays release 2025 🎁 | Gleam programming languageNews post: Gleam v1.14.0 released 07422
Adam Wyłuda @adamwyluda.com · 22/12/2025I’m no expert, but isn’t multilayer perceptron network part of the transformer architecture? 010
Reposted by Adam WyłudaMatthew Donald 📖🦖🤣 @matthewdonald64.bsky.social · 30/11/2025It takes 0.04 liters of water to make a single AI image. Meanwhile, it takes 5 liters to make a single piece of paper, and presumably even more when you add a pencil to it. Using AI to make art is literally better for the environment than pencil and paper. www.theatlantic.com/technology/a...theatlantic.comIt Takes More Than 3 Gallons of Water to Make a Single Sheet of Paper... and more mind-boggling stats that hint at a Waterworld future 1254
Reposted by Adam WyłudaAndy Masley @andymasley.bsky.social · 12/11/2025I shared a short succinct summary of my core strongest arguments that using chatbots is definitely not bad for the environment you can share with skeptical people in your life andymasley.substack.com/p/a-short-su...andymasley.substack.comA short summary of my argument that using ChatGPT isn't bad for the environmentTo share with anyone still worried 1154
Reposted by Adam WyłudaSimon Willison @simonwillison.net · 07/11/2025I have a hunch that current LLMs might make it easier to launch a brand new programming language, provided you can describe it in a few thousand tokens and ship it with a compiler and linter that coding agents can use simonwillison.net/2025/Nov/7/l...simonwillison.netCould LLMs encourage new programming languages?My hunch is that existing LLMs make it easier to build a new programming language in a way that captures new developers. Most programming languages are similar enough to existing … 17813
Adam Wyłuda @adamwyluda.com · 04/11/2025A program adjusting its weights by looking at something is also arguably not infringement, but a fair use. 181
Adam Wyłuda @adamwyluda.com · 01/11/2025Gen AI is when ML has any output. So the only useful ML is the one that does nothing? 071
Reposted by Adam WyłudaEthan Mollick @emollick.bsky.social · 28/10/2025New data on the corporate ROI from generative AI from a large-scale tracking survey by my colleagues at Wharton. They found that 75% already have a positive return on investment from AI, less than 5% negative. Also 46% of businesses leaders use AI daily. knowledge.wharton.upenn.edu/special-repo... 47515
Reposted by Adam WyłudaLaurieWired @lauriewired.bsky.social · 17/10/2025You’re (probably) measuring application performance wrong. Humans have a strong bias for throughput. "I can handle X requests per second." Real capacity engineers use response-time curves. 3447
Reposted by Adam WyłudaEthan Mollick @emollick.bsky.social · 17/10/2025I hope as we move past the first wave of AI criticism ("it doesn't work, all hype") we get a new wave of AI criticism rooted in the acknowledgement that, yes, these systems are very powerful & quite useful and focusing a deep exploration of when AI uses are uplifting and when they are detrimental. 711920
Adam Wyłuda @adamwyluda.com · 09/10/2025I’d also argue that being anti-AI is gatekeeping, for example people whose English is not the first language or aren’t great at writing, LLMs can help share their knowledge with the world. (Edited 7:32) 000
Reposted by Adam Wyłudadame @dame.is · 19/06/2025AI/LLMs are a major accessibility technology — being against this technology is to be against advancing human accessibility, and i think there’s a case to be made that it’s borderline ableist 7676
Adam Wyłuda @adamwyluda.com · 07/10/2025And with VAT, the B2B transactions are exempt from it, it’s only paid once by the end consumer, so it has little effect on overall economy. Meanwhile tariffs are a turnover tax, which compound each time a good passes the border. 000
Reposted by Adam Wyłuda⚡️🌙 @dystopiabreaker.xyz · 06/10/2025if you're curious about the architecture and mechanics of LLMs, this site has a really excellent explorable interactive visualization. it helps build intuition for how massive these models are, what 'interpretability' means, and the complexity involved here bbycroft.net/llm 1015923
Adam Wyłuda @adamwyluda.com · 05/10/2025Saying that we already know everything about LLMs because we know how they work on the lowest level is like saying that we know everything about mathematics by just defining axioms, or that we have computed everything just by inventing CPU. 070
Reposted by Adam WyłudaTangled @tangled.org · 03/10/2025hello tanglers! this october, we want to get more people into making their first ever open source contribution! to help them out, we've got the 🟣 'good-first-issue' label! head to tangled.org/goodfirstissues to see a list of issues across different repos that you can contribute to! 🤓 613324
