Your curated collection of saved posts and media
@aiNeoHuman Facts suggest it’s one of the fastest growing by GDP. https://t.co/KF56Knqd45
Even understanding the history of technology can help. For example, Pinch & Bijker’s studies of the early 🚴♀️ showed how tech is often not deterministic, its path is shaped by the social groups that use it & their view of its problems. Some useful lessons. https://t.co/CxVGOAJcOh https://t.co/r1UZCqcBS5
Increasingly, the findings of controlled experiments go much further: doctors with off-the-shelf AI outperform those without in diagnostics… but AI alone outperforms doctors. Harder to know what to do with that. What systems or interfaces will result in better human-AI teams? https://t.co/VoWdbPLOxc
Almost every study shows doctors with AI perform better than those without. Now AI is achieving perfect scores in medical licensing exams. You will simply expect every professional services provider you go to will use AI in the future or you won’t trust the advice.
Bing hung up on me all the time in 2023 if I antagonized it too much. https://t.co/rgUQXa8zdq
As part of our exploratory work on potential model welfare, we recently gave Claude Opus 4 and 4.1 the ability to end a rare subset of conversations on https://t.co/uLbS2JNczH. https://t.co/O6WIc7b9Jp
Looking at the ARC-AGI benchmark is a useful way of understanding AI progress. There are two goals in AI, minimize cost (which is also roughly environmental impact of use) & maximize ability. It is clear you can win one goal by losing the other, GPT-5 seems to be a gain on both. https://t.co/JosfuAzfdZ
The odd bit is that GPT-5 is two things, a router and a model name. So when you use “GPT-5” in ChatGPT you might get any of these models (maybe not the very lowest, but we don’t know for sure), as OpenAI optimizes cost versus presumed model ability to answer a question. https://t.co/k349e3CwJC
“Demonstrate recursion in a paragraph. Be very clever.” You can see a couple of models are very much “coder brained” (GPT-5 Pro and Grok). https://t.co/cinU5SNbku
Sorry, posted the wrong Gemini 2.5 Deep Think (I experimented with adding “this is not about coding” to the prompt, which didn’t change much, and just made GPT-5 Pro talk even more about coding) https://t.co/kNbMZw6Bcy
GPT-5 Pro, Gemini Deep Think, Claude 4.1 Opus, and Grok 4: “Write a two paragraph original secret history, smart and meaningful. Think Pynchon, Borges, Powers, or Eco for inspiration.” LLMs shine as connection machines (though some are better at writing this sort of fiction) https://t.co/JNFGn465Ic
I gave Claude’s conspiracy to Gemini Deep Research to fact check. It did really well. I then gave it to GPT-5 Pro; which had some suggestions to make the story better (also quite good). https://t.co/Df1gihT6Tc
People asked for kimi. Some interesting style, but did not get the assignment. https://t.co/IGKFxf4BxB
Microsoft, $MSFT, has said that these are the 40 jobs most at risk by AI https://t.co/381NuEJ8r1
Distribution unlocked: $NATION is now tradable on @coinbase app through @base DEX integration. https://t.co/Rc0zaSisDX
@apoorvasriniva https://t.co/BLRHLufn3a
torchtune was deprecated last month and a new repo for post-training at scale is being developed at Facebook. Hoping to hear more about this new library soon! https://t.co/Xn00gbWHfy
tell me again about how locked in you are https://t.co/GruwNVWJFU
@willccbb @PrimeIntellect You join Prime Intellect? Wow. I didn't know that. You're telling me now for the first time. https://t.co/hr5h886kJB
Thyme: Think Beyond Images "Thyme transcends traditional "thinking with images" paradigms by autonomously generating and executing diverse image processing and computational operations through executable code, significantly enhancing performance on high-resolution perception and complex reasoning tasks. Leveraging a novel two-stage training strategy that combines supervised fine-tuning with reinforcement learning and empowered by the innovative GRPO-ATS algorithm, Thyme achieves a sophisticated balance between reasoning exploration and code execution precision."
github: https://t.co/yXjHAGMOt5 huggingface: https://t.co/NwbyWsLYs5 project page: https://t.co/GA2qwW9wxu abs: https://t.co/sOcIpGXMpH
BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining "we introduce BeyondWeb, a synthetic data generation framework that produces high-quality synthetic data for pretraining. BeyondWeb significantly extends the capabilities of traditional web-scale datasets, outperforming state-of-the-art synthetic pretraining datasets such as Cosmopedia and Nemotron-CC's high-quality synthetic subset (Nemotron-Synth) by up to 5.1 percentage points (pp) and 2.6pp, respectively, when averaged across a suite of 14 benchmark evaluations. It delivers up to 7.7x faster training than open web data and 2.7x faster than Nemotron-Synth. Remarkably, a 3B model trained for 180B tokens on BeyondWeb outperforms an 8B model trained for the same token budget on Cosmopedia."
Supply chain customers are seeking alternatives to traditional solutions. To address this need, Vuzix just introduced the LX1 smart glasses for warehouse operations with hybrid voice and vision picking. The warehouse of tomorrow is here. Read via @Auganix: https://t.co/OnF6mNlJN0 $VUZI #SmartGlasses #AR #Logistics #SupplyChain #Warehouse #Tech
@Chad_Hurley @Saif_rogue177 Not AI either… https://t.co/fn0uZFlQpA
Public speaking isn’t just for the stage. It’s in meetings, videos, podcasts, team updates and even job interviews. If that sounds like you, I’m building Pause, a mindful speaking coach with AI feedback. Anyone interested to test and provide feedback 👉 https://t.co/oUe7M8j92F https://t.co/x5lQaIOhML
Seedance Pro UNLIMITED generations unlocked for ALL Pro users in Higgsfield! Bar is raised: 720p quality, up from 480p. Cinematic AI video, smoother than ever. With 30+ built-in presets, no prompts needed. Welcome to the Click-to-Video world! Retweet to get a guide in DM. https://t.co/aoWvAeiO4R
Why not a TopBar? Introducing a new custom component for @Gradio: a sleek, collapsible, and centered TopBar for your app's main controls ✅ Customizable height & width ✅ Stays centered ✅ Floats above other elements Give your Gradio UI a new look! https://t.co/ojddiMYWYD #gradio https://t.co/U8B9OqBJk5
Wan2.2-TI2V-5B @Alibaba_Wan is now available in anycoder for text to video generation for your vibe coding apps using @replicate on Hugging Face inference providers https://t.co/nyZJcqdoRp
app: https://t.co/esPDyHE1YC
testing a new workflow for anycoder, chat directly with commands instead of any toggles/buttons anycoder coffee, text to video: coffee pouring into cup https://t.co/Z91mzv1Wmx
this is now live, check beta: chat UI to use https://t.co/4L4CJWi5nR
testing a new workflow for anycoder, chat directly with commands instead of any toggles/buttons anycoder coffee, text to video: coffee pouring into cup https://t.co/Z91mzv1Wmx
Since everyone loved my Chinese lab ranking based on open model releases, here’s the equivalent for the Western companies based on released open models aligned in tiers to comparable Chinese companies. We have no one comparable in the top two tiers. https://t.co/jzZlCk6Mug
A tier list of China's top 19 open model builders. Who did we miss? At the frontier * DeepSeek * Qwen Close competitors * Moonshot AI (Kimi) * Zhipu / Z AI Noteworthy * StepFun * Tencent (Hunyuan) * RedNote (Xiaohongshu) * MiniMax * OpenGVLab / InternLM * Skywork On the rise
From Dan Wang’s new book. Interesting thesis. Engineers do tend to think in terms of solving problems. Lawyers profit when there are unsolvable problems. 🇨🇳 https://t.co/aPlie8vZPW
These 302 people have just had their social credit score slashed. If you are one of these people, and discover you now can't find work, join clubs, or book plane tickets -- remember, you brought this on yourself. https://t.co/JcgXu5EX7R
The best piece of often overlooked advice that I got at rentech (by Jim himself) in the earlier days was : “Always remember, first sort, then regress”. We lived by this rule. If you have data (X,Y) sort X and Y independently, this gives a better fit.
