All entries

Open models should compete on cost and specialization, not frontier benchmarks @natolambert puts it well: the right benchmark is savings in compute and time, especially for repetitive agent tasks deep

open-weight LLMs have come a long way on agent tasks! but the harness you wrap them in matters just as much as the model itself, and arguably the interface you use to drive that harness matters even m

This post by AI researcher Gary Marcus summarizes major events in OpenAI's recent history across three acts: ChatGPT's record-breaking success, the dramatic firing and rehiring of CEO Sam Altman, and

People ask me what my AI agent team does. The reality is that it is incredibly horizontal, incredibly open and flexible, and fully driven by a flywheel of verbalized, documented, and reviewed goals-co

Perplexity and Computer now connect to premium health sources, starting with NEJM and BMJ Group, with 9 more medical journals and clinical databases on the way. Ask health questions and get answers ci

arXiv:2512.07578v3 Announce Type: replace-cross Abstract: Global SHAP explanations are typically presented as feature-importance rankings, which identify variables that matter to a black-box model but

Pinecone Marketplace, live today. Until now, building an internal app that answers questions over a company's own docs, contracts, tickets, and wikis meant months of retrieval pipeline work: embedding

Sam Altman expresses enthusiasm about the advancement of voice AI models and their potential to improve human-computer interaction. He notes the emerging trend of users already adapting their communic

Proud of my analyst, Hermes ​ This is what he has achieved after 1.5+ months of internship with me ​ - Got me off X (doomscrolling cut from 2 hrs → 30 mins) - Practically paid for his salary via PMs b

.@Redisinc is returning to Interrupt! Say hello to their team at the expo hall to learn about their fast memory layer for chatbots and AI agents as well as their ready-to-use tools for building AI app

Run an entire company with agents 🔥 It's always awesome to see companies continuing to innovate on AI-native UI/UX, particularly around multi-agent coordination, to solve deeply complex tasks beyond w

Security remediation is an engineering capacity problem. AI has collapsed the time to exploit, but defensive tools haven’t kept up. Today we’re introducing Devin for Security: a set of workflows for r

See you all soon! We've got some fun announcements ahead. I'll also be doing a workshop on 'how we Claude Code' with some workflows I'm excited to share. Don't worry if you're not there, everything wi

Seedance 2.0 is an update or feature release for ComfyUI, a node-based UI for Stable Diffusion and other generative AI models. This announcement likely describes new functionality, improvements, or en

// Skills as Verifiable Artifacts // Pay attention to this one, AI devs. If you ship agent skills, your runtime is treating signed-and-cleared skills as trusted by default. This paper argues a skill i

So let me make sure I’ve got this correct: >Google gives Anthropic cash to pay Google for compute. >anthropic gives compute away for a loss, and this discounted compute draws in large user bases to ca

so well explained ! this is like applying Richard Sutton’s Bitter Lesson to agent systems. leverage comes from feedback loops, not handcrafted prompts. attaching feedback to traces is the missing piec

Some things never change. If you don’t understand this one, you don’t understand what’s happening AI. Marcus, 1998: neural nets have trouble generalizing far beyond the data. Marcus, 2001, 2012, 2019,

Perplexity offers a sourcing screen feature that allows users to input target criteria and receive a filtered list of companies matching those specifications. The tool provides transparency by explain

Sure looks like Bun is at least exploring a port from Zig to Rust given this docs/PORTING.md guide for coding agents https://github.com/oven-sh/bun/blob/46d3bc29f270fa881dd5730ef1549e88407701a5/docs/P

Thank you @UN Secretary General @antonioguterres for an excellent discussion on the current state of AI capabilities, and the urgent need for robust global governance. International collaboration is c

The brain does not predict every upcoming word during language comprehension with maximum precision, contradicting the assumption that next-word prediction serves as the central computational objectiv

the cofounder 2 launch by @intelligenceco hits hard for me especially because this was the prompt i tried 3 years ago that turned into babyagi @jennyai I’m about 30 prompts/iterations in from this sta

FLOPs (floating-point operations) serve as a practical metric for measuring AI computational capacity because they have a precise technical definition, can be directly priced and measured, and offer a

The crux of the Musk-OpenAI trial so far is that Brockman keeps acting as if selling chatbots and API access for profit (which is what they do now) is exactly the same as the original mission (which w

Gary Marcus highlights a concerning trend where non-technical teams are directly deploying production code, which raises significant risks around quality control, security, and system reliability in f

Sam Altman, CEO of OpenAI, recommended trying ChatGPT's new instant model, suggesting it performs significantly better than thinking-model-only approaches. The post encourages users who have relied ex

The unreasonable effectiveness of LLMs is what makes them so weird. The labs don’t need to decide what kind of AI to build, because better LLMs do better at most things. Finance? Pig disease identific

There’s a bunch of conflicting stances I don’t fully understand in the debate of Proprietary RL’d vs Open Harness, Model intelligence, and Agent Labs building harnesses for bespoke tasks Not all of th

Agent observability is enhanced by storing feedback alongside execution traces, transforming them from simple logs into a learning system that enables continuous improvement. This approach allows deve

Eridale is an initiative by Yohei Nakajima that combines AI and robotics to address universal food accessibility and improve dietary quality. The project aims to tackle malnutrition and food inequalit

Today, the CEO of @coinbase announced a 14% headcount reduction. Let's talk about why. He cites AI as a core reason for the shift, alongside market dynamics - though we don't know whether each of thos

Today we're launching Perplexity Computer for Professional Finance. Finance teams can bring licensed data from providers like Morningstar, PitchBook, Daloopa, and Carbon Arc into Computer. We’ve also

Too early to tell how the trial ends, but if Elon does win, the takeaway might be in part that Greg Brockman made two immense blunders: 1. Writing down all that stuff in his diary 2. Blowing off Elon’

Too much of the solutions & language for agentic systems comes from a coding perspective (control planes, hooks, loops), but I really think that the field of management and organizations can tell us m

toodles from mickey mouse clubhouse was weirdly ahead of its time wake phrase: mickey mouse clubhouse launched in 2006, and “oh toodles” trained toddlers on the assistant wake phrase years before siri

Topaz Astra 2 is live. The sharpest cloud-based video upscaler in the industry is now available via partner nodes. Adjustable levels, prompt-based control, and next-gen detail for everything from land

'Traces everywhere. Feedback loop? Nowhere' @hwchase17 This is the part that gets skipped. Traces everywhere. Feedback loop? Nowhere. Building agents for clients and the hardest sell is 'we need to lo

ComfyUI announced a cloud-based template for API integration with Luma AI's unified image creation model, accessible through their Comfy Cloud platform. The template (api_luma_uni_1_image_create) enab

Comfy Cloud is offering an online interface to try ComfyUI's API with the Topaz Astra2 template pre-configured, allowing users to test the model without local installation. This represents Comfy's eff

Elon Musk announced the availability of Tesla, Starlink, and auto-translation features on the X platform, expanding the social media service's integration with his companies' products and services. Th

Cursor offers pre-built automation templates available in their marketplace that enable users to automatically investigate and fix continuous integration (CI) issues, reducing manual debugging time. T

Vibe coded video games are starting to get... actually fun?! @PlayAstrocade's user growth is proof in point. Try out Sequoia Capital: The Video Game :) Five months ago, @PlayAstrocade users were playi

Want to run the same harness across multiple interfaces? Try ACP. Deep Agents ships with it out of the box. open-weight LLMs have come a long way on agent tasks! but the harness you wrap them in matte

We cut model cold-start times from minutes to seconds. 60x faster, saving thousands of GPU-minutes and hundreds of terabytes of transfers every day. It turns out GPUs already holding weights are faste

we have very efficient models, especially for their capability level happy codexing yo, i'm actually worried. codex limits are genuinely insane so it's sus af .. i feel this is an intentional move for