2026-11-03 に公開されたAI Pulseの更新一覧です。
82 件のシグナルThings will change again and again and again. We’re still very early and agents today are incredible, and yet so frustrating still. Can’t wait to crack things to get that flow state back. Long journey ahead.
I rebuilt OpenCode to run inside browser devtools with console.log statements! I hope you (errr, your agents) will build some amazing console apps, too: https:// textjs.org
Some people say that Opus 5.5 can already completely recreate OpenAI's new promotional video using code. No video generation models were used, just Three.js, GLSL, and an SDF female lead. Next time, we can try a tougher
We got drunk, going from a world where only certain specialists can build things to a world where anyone can build anything. Now the hangover is setting in. Just because you can, does not always mean you should
Every quick LGTM on an agent-written PR is a small bet. Stop making them.
I think in the AI era, we can put more energy into product details, like making the user experience better, performance better, and ensuring safety is no issue. As for code details, they're not that important, because Ag
AI made writing code dramatically faster. Running that code in production didn’t get any easier. Traversal gives AI the production context coding agents lack, then feeds what it learns from incidents back into those agen
Big news for our developers! We are shipping a developer preview that dramatically expands the capabilities of all apps to integrate into Siri. So excited to see what you folks do with this.
Simantic is the simulation layer that lets engineers and AI agents test and validate firmware at the speed of software. Frontier AI models have made generating firmware code cheap and effortless, but testing still depend
guys i'm losing my mind here. almost every reply i get is AI. every message request i have is AI. are there humans left here?
We looked at the data and the data stared right back at us 🥀 Some insights into AutomationBench, OfficeQA Pro v2, web access in public benches and why monitoring is hard. This has been in the work for months, more to co
LangSmith Bring Your Own Cloud (BYOC) now runs on @AWScloud , @Azure , + @GoogleCloud . Your traces and agent deployments stay in your cloud account while we run the data plane. ☁️ GA on AWS, Beta on Azure + Google Cloud
Baseten’s revenue grew 20x in a year, Silicon Valley’s inference cloud valuation is about to double to $30 billion—why? In January this year, it was valued at $5 billion; by June, $13 billion; the new round in talks is f
I'm really excited about this project that we've been working on: mlx-learn is here! Apple silicon accelerated "classical" ML algos. Up to 30X speedup for training, and 300X speedup for inference! Same API and matching r
Friday was my last day at Anthropic. Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env. I steered it for tens of hours but never manually wrote a line of code. It worked, I was hook
anyone who is great at deeply explaining things now can make 1h long videos explaining any deeply complex topics in a highly intuitive way like someone could produce a 1h video about how markets work and how things get c
Use AI to analyze your portfolio on Interactive Brokers – connect your investments to ChatGPT, Claude, and more. Paid post with @IBKR Ad
Claude Code 2.1.318 is about to be released #cccnext
One prompt. One 30-second take. Eight scenes, zero cuts. The subject stays locked dead centre while every background changes on the beat. Seedance 2.5 on OpenClips. Sign up FREE.
There is a version of the future which is rebels with hyper advanced AI vs monolithic incumbents with boatloads of free cash flow and reg capture…
Today, we’re excited to launch Prism into General Availability. Prism is the AI platform for hardware engineering. Describe what you want to build, and agents design the CAD, run the simulations, and write the report. Fr
i made AnyScene to build any scene in an hour. i built AnyShot to render those scenes in a few minutes. on my RTX 5070 Ti, a single walkthrough like this takes 6 hours to render. AnyShot fans it out across up to 100 RTX
Everyone is talking about agents now. @AidenChaoyangHe was already building for this moment before it became obvious. I still remember him describing the shift from isolated assistants to team agents—and the Agentic Inte
Gemini contacts your vendors and places restock orders. Now your shelves are ready for every rush. Make work less work.
Teamily AI @Teamily_AI · 8h Article The Future Is the Agentic Internet: Moving Beyond Personal Agents to Team Agents Dear Teamily AI Customers, Ecosystem Partners, and Investors, Over the past few months, Teamily AI has
A lot of content engineering challenges have pretty standard solutions, which has lead to some engineers saying agent engineering is solved/boring True to some extent, but there are a bunch of (fun) new ones emerging as
The harness has started building itself. Singularity reached!
⚡ Inside the 2026 SOTA of Software Factories with @zachlloydtweets of @warpdotdev ! Warp's open source cloud factory approach has been resonating bigly this year, leading to multiple top @aiDotEngineer talks and an invit
这么多新功能，期待Teamily AI 3.0！
Should you rely on public benchmark scores when choosing a model? @VikParuchuri argues "no" in this talk. To illustrate this, he shows specific problems found in popular OCR benchmarks: ParseBench: • A test expects 310 g
The agent-native financial stack. Bank accounts, cards, on/off-ramp and more. Setup with a single prompt.
OpenAI's next-generation large model might be coming soon, perfectly following up on what Altman said on November 2 in Abu Dhabi: "The next GPT will be released very soon." However, this is just speculation for now, with
Sandboxes have no moat. Just fat margins. We're launching https:// agent37.com to kill those margins. For example, a sandbox with 2 vCPUs and 4 GB RAM when left running for an entire month: Vercel: $249 Cloudflare: $145
You can't claim both that AI is fake and that it's destroying jobs. If it were fake it wouldn't be able to.
The AI bubble bursting is going to be one of the biggest reality checks in tech history. These companies have spent years convincing us that AI is the future while burning billions, destroying jobs, draining natural reso
Josh Gruenstein of Tutor Intelligence on how to value a robot: "You take how much work a robot can do and you divide it by its cost, and that's it. Everything else is commentary." If robots can do that work for less, he
Applied Compute update: - crossed $100M annualized revenue run rate up from zero 18 months ago - raised $350M at $3.25B led by Elad Gil - 40% of customers use AC for both custom model training + production inference - i
BlueChip. Built in America. America made the first electronic computer to win a war. Now we depend on our greatest adversaries to build the computers that run our economy and defend our nation. Today, we bring it home.
agentic browser requests jumped ~5x in september alone... browser use is becoming a key primitive for agents! icymi, we shipped it in (managed) deep agents last week now your agent can do anything interactive on the web!
Robot control has seemed reserved for frontier LLMs like Astra and Opus 5.5. Our results challenge that view: robot-use is a general capability emerged across many LLMs, including open-weight models like Qwen and DeepSee
Browser use lets agents act on sites built for people, and it puts real pressure on the context window. Here's how we chose and tuned the browser tools in Managed Deep Agents, built on @Stagehanddev from @browserbase
You can now create and share your avatar in Gemini to star in images and videos with friends. Share your personal avatars, tag each other in prompts, and maintain character and voice consistency. Plus, you're in control
Introducing Vyra! We're building the personal video editing agent for shortform content. Just upload your raw footage, prompt, and Vyra edits it for you. Try now at https:// usevyra.com
It looks like Astra didn't natively read Qwen's weights? As I understand it, Qwen would get stuck in a loop and not answer the question. Astra used code to perform some standard interp techniques, found a looping directi
RL can help collect good data, whether you're doing BC or RL. It can also stack two cubes on top of another cube. Fun paper with @qiyang_li @zhiyuan_zhou_ @CharlesXu0124 on how RL can assist with robot data collection!
High utility bills and data centers are making public service commission races more competitive in some states
Today we’re launching Sutrix (YC F26): AI software that turns patient records into study-ready datasets for healthcare research teams. Across 10+ institutions, we’ve cut data preparation time by 80%, helping teams build
Blacksmith (W24) is building a faster cloud for continuous integration (CI), helping developers test and ship code as AI coding tools generate more of it. As software production accelerates, testing and validation are be
Our picks from last week's model drops 🦙 → LFM3 by Liquid AI: small models for text, image and speech, with d1 decision models → Mistral Large 4: large, 1M context frontier → Flint by Recursive: mid-sized, built for cod
Google's open-source SDK lets TypeScript and JavaScript developers call Gemini models to generate text, images, and other content.
How will the UK compete for the world’s top AI talent? "I think we are now in a very compelling policy position." UK AI Minister @KanishkaNarayan tells @TomMackenzieTV at #BloombergTech. Watch the conversation here ⏯️ ht
Made two with @op7418's prompts — super kawaii
Claude API Public Beta Message Threads: No Need to Resend History for Long Conversations Each Round, Faster Replies, Same Pricing Anthropic today launched a public beta for "message threads" in the Messages API (the prim
Breaking: Google will turn off web search for schools that opt out of A.I. for kids.
Anthropic Hiring Communications Lead for Claude Developer Platform, Salary $265K to $295K Anthropic is hiring a "Claude Platform Communications Manager" to handle external communications for the Claude developer platform
DeepSeek Harness is way too hot. Opened two WeChat discussion groups in the evening, and they were both full with 200 people in less than 10 minutes. Based on past experience, the group gets pretty lively around 350 peop
Former New York Fed President Bill Dudley says we should be worried that the current AI boom may be sowing the seeds for a future financial crisis (via @opinion )
Founders send decks with no net revenue retention, no burn multiple, and a market slide that says $100B TAM. Then they wonder why no one replies. Run your deck through the SaaStr AI pitch deck review and see which number
AI moderators can now handle a full hour of open conversation with a real person on camera. @atypica_AI just introduced Atypica AI Moderator for face-to-face consumer interviews at scale, with many sessions running in pa
Now in OpenClaw - it was collaborating with them on this! We have a lot more updates to the model coming...
Can we forecast benchmark accuracy from models that score zero? We (co-leads @LukeBailey181 @rckpudi ) turn a benchmark into a series of easier hinted versions. Jointly modeling scale C and hint level H enables forecasti
Is the wheel thing because the company doesn't allow using Claude? Indeed, Astra isn't as handy for writing code as Opus 5.5. Good code is standard and doesn't need special explanation; it's the bad code that requires pe
New!! Jason's Takes on This Week's 20VC: Anthropic's S-1, Microsoft's 27% Buyout, and My 100-Day Rule The bar for 2027 raises: show your path to $1B.
New!! Why You Don't Want Humans Doing Outbound Anymore A rep pitched me with a 2024 sequence nobody had read since. At volume, humans stop reading what they send. Agents read every account, every time.
Grok @Bot 's new calling feature is very impressive. You just tell it who to call and why, and it handles the conversation for you. It shows a real-time transcript of the call, and if Grok needs additional info from you,
With AI, writing code has gotten way faster, but verification is still slow, especially waiting on GitHub Actions CI. I spent this afternoon giving the CI for Mole, Kaku and Pake a full overhaul, and the results were gre
AI Arms Race Turns into a $1.3 Trillion Data Center Leasing Gamble. Oracle, Microsoft, Meta, Amazon, and Google have committed to over $1.3 trillion in future leasing expenditures, with about 70% added in the most recent
Proud to be a launch partner in the @FireworksAI_HQ Frontier Ecosystem!
using astra as a system2 model is very interesting and also bitter lesson pilled! eagerly looking forward to reading on the approach here. congrats on being the only independent research team to make it to top 10!
ChatGPT and Codex are helping engineers build for the physical world. We’re sharing our progress across mechanical design, electrical engineering, and simulation—from helping our hardware teams iterate faster to testing
I'd like to tweak this a bit: Could you conveniently let me know the magpie version, which agent you're using to connect to Claude (is it Claude Code itself or something else), and when it gets stuck, is it that there's
Capital One researchers developed APT (Adversarially Pre-trained Transformer) for zero-shot tabular prediction. APT uses adversarial synthetic data agents and a mixture block architecture to handle arbitrary class counts
After using the Claude CLI connection with magpie, it gets stuck particularly easily and doesn't output anything, so I had to switch back to using the CLI directly with herdr.
Suddenly had an idea to build a global note-taking tool for Obsidian. Input prompt: "Help me develop a Mac software that launches via a shortcut key (customizable key combination), allows inputting content, dragging in i
I replicated this with Claude Code! It took a live webcam 7 km south of Southampton, UK and predicted its location purely from doppler effect in audio spectrogram + open-source aircraft location data. Lo and behold, the
Its very interesting to note the pace of application features from the foundational model companies. As the cost of the marginal token approaches zero, the value is not in selling tokens. It is very clearly in moving up
Fable 5.5's video-making capabilities have been further improved, it's genuinely impressive—being able to achieve this level with pure code. But for regular folks, this capability boost is pretty limited. I'm more eager
Hugging Face reposted Sergio Paniego @SergioPaniego · 6h Article MOPD: how frontier labs merge RL experts into one model and how to do it using TRL TRL now supports MOPD. AsyncDistillationTrainer routes each prompt to it
Users can now auto-connect plugins when they log in to Claude, with MCP servers that accept Claude as a trusted issuer. If you’re building plugins for Claude, add support so users get started faster and stay connected wi
Building with AI agents at work? Join OpenClaw Enterprise’s bi-weekly office hours. Bring your questions, share what you’re building, and help shape an open-source, vendor-neutral platform for enterprise agents. Register
Agent keeps filling up my computer at the drop of a hat—what's the best macOS virtual machine management software?
My next mission is to direct tokens at useful and meaningful problems in life 💪 Need more tokens!!! Anyone know how to get frontier level tokens for open-source projects? 😅