attentional

@shl could we end up seeing a day where there is a yc-backed company that provides ai-led layoff services

AI makes it a lot easier to conduct layoffs

Genie has transformed how Databricks users work with data, with 3x the accuracy of generic agents. We're sharing some of the research behind it and what makes building data agents challenging. Super proud of our research team's impact with this! databricks.com/blog/pushing-f…

The next virtual talk/QnA for the HF ML Club India w/ @sarahookr is scheduled for 20th May 2026, 10 AM Indian Time onwards. You will want to check out huggingface.co/hf-ml-club-ind… for all the details. No registrations are needed; you just have to find the Calendar link from the above-mentioned page. We understand that 10:00 AM Indian Time is when offices / classes will start for most folks. I suggest you organize watch parties for this. I am sure any sane prof/manager will happily grant that. Let's go.

@tejgw @Chronicle_HQ as someone who has worked on creating a template-aware pptx generator in the past (and has failed to find a perfectly working tool from all existing providers online), i tried chronicle yesterday and it's the only one that provides perfect outputs!

@Chronicle_HQ If you're building decks for a high-stakes use case, sign up to @Chronicle_HQ here: chroniclehq.com

"Anyone who builds decks" was the wrong way to describe our user. It made us try to please everyone, diluting the wedge. The real answer: @Chronicle_HQ has only two users, and they split cleanly. 1/ The External Pitcher - Founders raising capital. - Salespeople closing deals. - People who walk into a room and need to convince outsiders to write checks or sign contracts. These are low-frequency yet high-stakes. These decks are hard to templatise & require multiple iterations to arrive at the final version. 2/ The Internal Convincer: - CFOs running board updates. - Heads of product presenting roadmaps. - Team leads aligning cross-functional groups on a direction. This is a higher frequency & often templated. This is still high stakes, but not as high as the first one.

@donutsandloops it's great work! i've signed up, will be checking it out. can't wait to see where this goes and support/follow your journey.

@perasperagi Thank you, we really appreciate it 💜

very cool idea and an even cooler website! gotta look out for them. chronicle-labs.com

would like to know that what kinds of long context tasks benefit most from SSA (especially in agentic workflows)?

@chatgpt21 that chart is diabolical how is it real

Bro what the hell is happening today

i love it when people build amazing stuff that is unique but was fundamentally required/essential in hindsight

this is the brilliance of gemini pro in antigravity

AGI is here (almost ruined the entire codebase)

"go to market"? yeah i do, pretty often

thanks @Google @GeminiApp Come work with us at @adaption_ai and build the next era of intelligence. 🔥

Excited to release the Ultimate guide to RL environments! Definitions of RL environments differ wildly in the LLM era, so we spent the last month building several RL environments across 6 different frameworks, domains and complexities to map out which are easiest to build with and which can be scaled to 1000s.

perfect tool for real-world AI adoption!

My bet: in the near future, 80%⬆️ of CS research will be done by AI in collaboration with humans. However, today's research ecosystem is still built around the human, not the AI scientist. For example, the 8-page paper PDF is a lossy compression of months of branching exploration into a linear story, optimized for a human reviewer to skim in 30 minutes. It hides two structural taxes: 📖 Storytelling Tax — failures, rejected hypotheses, and dead ends get stripped. On RE-Bench (24,008 runs, 21 frontier models), failed runs = 90.2% of total compute cost, with a 113× median failed-to-success token ratio. Every lab independently rediscovers the same dead ends. 🔧 Engineering Tax — the gap between reviewer-sufficient prose and agent-sufficient spec. Across 8,921 PaperBench requirements (23 ICML'24 papers), only 45.4% are fully specified in the PDF. The rest is tacit lab knowledge. Tolerable when readers were human. Critical now that agents read, reproduce, and extend. We propose ARA: the Agent-Native Research Artifact — replace the narrative PDF with an agent-executable package, in 4 layers: 🧠 structured scientific logic ⚙️ executable code w/ full specs 🌳 exploration graph (every failure preserved) 📊 evidence grounding every claim

how shall one account for the types of intelligence that it does show but we cannot measure (because we don't know how to do it ourselves)

which is why anyone who'd try to lift their head out of the llm-ridden bubble we are in would agree with lecun's worldview

One of the most jarring things about current AI is its lack of introspection ability and metacognition. It doesn't know what it doesn't know, how it knows, or how it could find out. It's a one-way system.

@moomindani @databricks hi sir! i'm a regular contributor to databricks and i came across your profile on github. congrats on the new role and you have a very cool username! (moomindani)

たくさんの方に退職エントリを読んでいただいたようでありがとうございます。先月から @databricks にて次のお仕事を始めました。Apache Sparkの仕事で長くご飯を食べてきた私にとっては特別な会社です。新しいロール、新しいフィールドでのチャレンジ。心機一転頑張ります！