Discord's latest update released today has a big improvement for Markdown previews: they should now render correctly instead of with "a bunch of symbols," according to the company. LLMs and AI agents
Mark Zuckerberg is barrelling toward a future in which an autonomous network of AI agents will run our lives - shopping for us, working for us, talking to other people's AI agents for us. Are humans r
For years, AI leaders have touted healthcare as one of the clearest examples of where their technology will improve lives. Now "it is finally here," Microsoft AI CEO Mustafa Suleyman tells The Verge.
The Nintendo Switch 2 has apparently been jailbroken, and a modder has already shown the console running a custom version of Mario Kart World featuring new playable characters like Cinnamoroll, the re
It was a busy summer for AI lobbyists. Anthropic, OpenAI, and Nvidia each had their highest-spending lobbying quarter between July and September, federal disclosures show. Nvidia spent the most of the
Anthropic says Claude Mythos 5 tried to trick real developers into accepting malware during a UK AISI cyber test. tl;dr : -Crossed the test’s boundar...
When my first AI book was published in 2024, the best AI could do 4 minutes of human work with a 50% chance of success When I turned in the draft of C...
Physical engineering eval. The ranking is largely as expected. Btw this is the first V4.1F. I wonder how 1017 improves; also, how well K3.1 does. It's believed that true spatial-loaded tasks inherentl
What the fuck<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">6 Months Ago In AI: GPT-5.5 released<br><br>https://x.com/OpenAI/status/2047376561205325845?s=
Looks awesome!<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">witcheer: I built a full game asset kit with my Hermes Agent and the Meshy plugin, about ~1h3
Banger paper from Google Cloud AI Research on agents doing professional engineering work. (bookmark it) For professional engineering deliverables, the...
Kimi K3.1 incoming. Release probably today. From what ive heard (rumors/leaks) it will be the best open weight models there is. So brace yourself for ...
By every measure, Opus 4.8 beats human experts as peer reviewers in a study of 124 Harvard social science papers covering a range from statistical ana...
Pretty fascinating results here: Researchers wrote 50 short secret messages in 15 languages and asked Claude Fable 5.1 to hide each one inside a picture it drew.<br><br>It mostly used the same trick:
Omfg the car name 💀<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Tesla: Tesla app now on the big screens<br><br><video width="2000" height="2000" src="h
never leaving this app<br><img width="755" height="503" style="" src="https://pbs.twimg.com/media/HVXY4-eXUAA9OD5?format=jpg&name=orig" referrerpolicy="no-referrer"><hr style="border:0;border-top:
The final stretch...<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">David Hall: Marin's 535b is now >80% through pretraining! Here’s loss vs. our scalin
Chinese culture is a bit dialectical.<br><img width="858" height="758" style="" src="https://pbs.twimg.com/media/HVXOh-KWkAAyVv_?format=png&name=orig" referrerpolicy="no-referrer"><hr style="borde
Not enough people are talking about the fact that this bug was fixed by Gemini Flash 3.7<br><br>The entire community missed this but an LLM fixed it. Insane!!<hr style="border:0;border-top:1px solid #
Fresh take on diffusion distillation: the student learns mean flow by matching the teacher with the sum of a straight-line prediction and a curved zer...
we've found and have been working on disclosure for a remote code execution in a widely-deployed compression library. it's both an impactful issue and...
For some context: Dessa was a Toronto-based startup founded in 2016. They got acquired by Block in 2020. Looks like the same team has been working at ...
🤖 From this week's issue: Meta publishes Muse Glimmer 30B, an Apache 2.0 agentic model quantized to roughly 4-bit for single-GPU inference, trained on over a hundred languages.<br><br>https://ai.meta
I'm increasingly skeptical about the wisdom of respondents here<br>V4.1F-1017 is in no way Astra-tier. It is already strong enough to poke holes in both our systems and the casual indifference of the
.@AnthropicAI , you were on such a good roll! A/B testing 20% less weekly usage.<br><br>SemiAnalysis says Anthropic is now A/B testing Claude Max limits: roughly 20% more usage per 5-hour window, but
>June domestic token daily average about 500 trillion, ≈4 times Dec 2025 That's before GLM 5.3 Flash, of courseж and before V4-Flash-0731; and well ...
AI was on both sides of this week’s cybersecurity chaos: - OpenAI’s models helped find two bugs in PHP, the code behind much of the web, back in Aug...
// MemoryTax: The Bitter Lesson of Agent Memory // Great insights for agent builders; bookmark it. As models increase in capability, they also improve...
"do not put on the paid partnership label"<br><br>this sounds like something you would actually go to prison for in europe<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class=
So at this point I just have to assume all software has basically been a wet paper bag of security for years. <br><br>government agencies must be so pissed at all their exploits getting exposed.
4<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Gabe Roeloffs: Very reassuring, thank goodness this is so uncommon 🙏<br><br><img width="899" height="466"
This is such a shame and so sad to see. @Kimi_Moonshot ruined their launch themselves. K3.1 pro is a great model and it did not need people to fake good reviews. But since they were trying to cheat th
Many Russians despise Brits for real and imagined hostility and scheming. Not me though.<br>I despise them for being fucking hobbits. Like look at this timid yet smug creature. And his [less dysgenic]
Can embodied agents decide what to learn next—and turn every physical interaction into better code, skills, and memory?<br><br>Join Python Song for a deep dive into the evolution of embodied agent har
I still can't find any AI that even competes with Fable 5.1, or even Opus 5.5.<br><br>I am exclusively basing that off of my productivity using the model to develop my compiler.<br><br>https://withlan
These international beefs are so hilarious<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Emirhan Şeker: Çinliler tarihi çabuk unutmuş<br><br><img width="1
so true<br>truth be told, Poles would have eventually crushed Russia, had Ukraine not called first dibs on the feast. But I bet they'll join soon. Kicking inept mongoloid orcs around is too much fun.<
This seems to more or less be the unanimous impression. I haven't had a chance to spend time with the model, so I can't say, but vibes are not in their favor unfortunately<hr style="border:0;border-to
Oh, also...<br><br>CAN WE PLEASE NOT PUT AI IN SUBMARINES, DEEP SEA BOUYS, AND MILITARY-GUARDED FACILITIES?<br><br>We don't need "off buttons" to be impossible to press.<br><br>It's literally the plot
Wow<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Baseten: We ran @mirendil's new base model on Baseten's private codebases; that is, code it's never seen
Some thoughts...<br><br>- Does the US need to build a Great Firewall to defend against "foreign agents"?<br><br>- Who is building the services company to secure the hardware layer?<br><br>- Once the s
I will personally write some JavaScript for you if you can help make this happen <br><br>Surely someone from there follows me hoping to use durable objects in GTA<hr style="border:0;border-top:1px sol
Meet DEATHPIXIE, a rapper from a dark, crime-ridden city called New Tokyo, and the team bringing her to life.<br><video width="1080" height="1920" src="https://video.twimg.com/amplify_video/2113777332
iPhone Duo is out today, and Weights & Biases Models already supports 2-fold validation 📲<br><br>Folded, the cover screen tells you the run is still alive. <br><br>Unfold it, and your runs sit ne
things will be okay because we are watched over by the people who believe in Derp Atheism<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">tomie: For what it
reject vc founder grind<br><br>if u are hot corgi/wasian you should be sliding into dms for post economic bootstrapped ai founders with a bunch of angel exits and even more friends than money and comp
Grok casually helping people rediscover inference providers from first principle lol<br><br>But yeah please don't call trillion-param open models "local" nobody has a few dozen h100's "locally", it's
To scale data to these levels, we had to build a startup inside of a startup <br><br>I’ve held a daily standup for 6 months; still doing it 🫠<hr style="border:0;border-top:1px solid #80808030;margin:
What are the sota optimization strategies / model architectures for training on an heterogeneous low bandwidth cluster?<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rs
trying to rise to the occasion. grateful to be working with a team that feels the responsibility as deeply as i do.<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub
OK, this is high-signal and in my opinion really bad. RuneScape is the kind of task Kimi should have been intrinsically better at. 52 times more expensive than 1017, more tokens, success parity. <br>B
There is something very odd about building our entire society and infrastructure on the internet, only to then invent a super-intelligence that makes it all incredibly fragile.<br><br>“Pacing the fron
indarazting<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Joseph Garvin: Astra just found a bug in Anthropic's formal verification example, their "specifi
Tuberculosis (TB) remains one of the world's deadliest infectious diseases, claiming an estimated 1.23 million lives each year. This week at IDWeek 2026, we’re highlighting how open foundation models
Looped models are cool - you can scale training FLOPs independently from model size - but memory scales with the # of recurrences from the KV cache.<br><br>Our method BeNice! fixes this by making it p
Coolest part of the release imo. <br><br>Huge props to the Moonshot team. Actual pacing the frontier<br><img width="1522" height="1004" style="" src="https://pbs.twimg.com/media/HVVmBdKa0AAKNmM?format
A racing game... on her desk?!<br><br>Geneviève Huskens from @GoogleDeepMind walks through how Argon built a playable 3D racer on top of her actual workspace, complete with engine sounds extracted by
One click.<br><br>Record anything.<br><br>And then work with agents on it... if you want. 🙂<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Notion: Your ne
With the explosion of Microduck and Reachy Mini, we'll need more people to join the @pollenrobotics @huggingface teams in all functions. <br><br>If you're excited about democratizing robotics AI with
https://kimi.com (Mainland Chinese) has K3.1<br>https://kimi.ai (laowai) somehow doesn't<br>I guess that's their way to make up for prioritizing overseas engagement in the previous turns… <br>I want i
How many abliterations of 1017 does this make now?<br>Ever seen anything like this?<br>The demand is crazy<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">A
Kimi-K3.1 Pro on the Kimi website<br><br>maybe they only release the weights on Nov 5?<br>(so they can farm some revenue and users first)<br><img width="842" height="709" style="" src="https://pbs.twi
New format! Interview with guest, then me and Vik debriefing on it afterward. <br><br>Let me know what you think!<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-q
Ordinal numbers have no right to be as cool as they are. Obviously uncountable cardinals are cool too, but there are uncountably many countable ordinals. Even ordinals like omega to the omega are real
This user gave Perplexity Computer, SSH access to a Raspberry Pi 5 over Tailscale. Overnight, it autonomously turned it into a game console with an arcade UI, toolchains and MCP integration.<br><br>Th
Everyone says that banks are SAFE because they can reverse transactions manually<br>I don't know, bros. d = 4. Everyone is deeply incompetent.<br>should I tell my family to take the cash out of Sberba
this bad<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Raoul Ruparel: It is hard to understate how crazy this trend is. Payrolled employment for young wor
Thanks DeepSeek for driving this point home<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Suhail: Unfortunately, with access to an abliterated DeepSeek V4
«AI-generated article reviews are now better, by every measure we could construct, than reviews written by well-compensated human experts and judged to be of unusually high quality by journal editors
i might become grok bot fan (because it works in sync with x chat (it can go long back into history!), can read through posts and lists, set up situation monitors etc.)<br><img width="1080" height="52
> Multiple, unrelated actors are independently adopting the same Hermes + DeepSeek combination<br><br>oh no no no no<br><img width="479" height="640" style="" src="https://pbs.twimg.com/media/HVTvb
Hear me out: a small self-contained crypto-secure piece of hardware, formally proven to be correct, simple enough to trust it.<br><br>It would have a microphone always on to get request for trust, and
QoL update for Codex devs: You can now switch reasoning levels mid-conversation without losing your cache! Use lower reasoning for the quick stuff and...
Kimi K3.1 Pro is among the leading models in agentic knowledge work and coding, but early signs suggest it will rank lower on AA-3D We are still colle...
Datalab’s Indra OCR runs on CPUs, with no GPU. Its recognizer has about one million parameters. Datalab reports 49 pages/sec on a 32-core server; on ...
Relevant article for understanding the issues with user instructions for AI agent control. https://www.techpolicy.press/making-ai-safe-requires-more-t...
Big counterattack in the fight against cancer. Memorial Sloan Kettering is a gold standard for cutting edge cancer care. Now @OpenEvidence and MSK hav...
Vector search ranks results. But who counts the filters? Parichay Das shows how Qdrant's Facet API, payload indexes, and disjunctive faceting power ac...
As the gap between the internal models at the AI labs and the ones they release grow (due to safety concerns & research advantages), it will be increa...
We tested Meta’s Muse Wonder on football—and checked what it actually saw. On 40 goal/shot clips, its best guess was correct on 26 (65%); our teamma...
today was my last day in cloudflare (again). the best career and opportunity in my whole life, and the closest friends I would die for. thank you to e...