Knowledge catalogue
One of the best @latentspacepod shows and Substack with @_lopopolo @OpenAI and @swyx focused on harness engineering, rethinking processes from first principles, and scaling with AI systems. https://su
Durable is an AI-powered business builder that helps entrepreneurs launch and manage businesses in minutes, serving more than 3 million businesses with a team of just 6 engineers and no DevOps st...
Waldium is a YC-backed, two-person agentic CMS that automates content research and creation, and gives every customer blog its own Model Context Protocol (MCP) server endpoint so AI agents can quer...
On April 10, 2026, Simon Willison posted a short note sharing a podcast clip from his ~1 hour 40 minute conversation with Lenny Rachitsky, highlighting a segment about kākāpō parrots. The kākāpō (*...
Vercel's AI builder v0 has acquired the team behind new.website, a tool founded to make creating beautiful, production-ready websites effortless through built-in primitives like forms, databases, S...
I was unable to retrieve the specific tweet content from that URL, as the post requires a logged-in X (Twitter) session to access, and search results did not surface the full text of that specific ...
SERHANT., an AI-native real estate brokerage, built its internal AI platform S.MPLE on Next.js and Vercel, scaling it from a 200-agent pilot to over 900 agents without replatforming or rebuilding i...
TGIF! Here are some of our favorite updates from the past week: — Notebooks in @GeminiApp, an integration with @NotebookLM that enables you to retrieve context from your private notebooks or convert y
This week @HumanXCo was a reminder that the AI conversation has moved on. No more benchmarks vs. bubble debate. The questions were practical: → How big do you need to be to train your own model? → How
After playing with it a bit, Meta's Muse Spark Thinking is fine so far, but really doesn't match the current Big Three models. It also is a bit... weird. Like some strange language & tone, a little lo
Can’t trust WhatsApp 🚨 WhatsApp’s “end-to-end encrypted” privacy is a total lie. New class-action lawsuit just dropped: Meta secretly let employees, contractors like Accenture, and third parties read,
The AI Engineer World's Fair (AIEWF) held its keynote livestream on YouTube, featuring prominent speakers including Malte Ubl (@cramforce, CTO at Vercel), Raia Hadsell (VP of Research at Google) ,...
Ryan Lopopolo ([@_lopopolo](https://x.com/_lopopolo)), a Member of the Technical Staff on OpenAI's Frontier Product Exploration team, hosted an AMA on Slido in conjunction with his appearance on th...
Amazon released the Nova 2 model family on December 2, 2025 at AWS re:Invent, comprising Nova 2 Lite, Nova 2 Pro (Preview), Nova 2 Omni, and Nova 2 Sonic, all featuring a 1-million-token context wi...
@swyx with all the flaws and these examples, is AI still net worth it? @aiDotEngineer @GergelyOrosz the leaderboards came from leadership pushing for usage individually we are moving a lot faster, as
The US frontier labs have all walked away from open weights. They continue to occasionally release excellent open models (Gemma 4, etc), but they are smaller models that are not competitive with their
WhatsApp’s “encryption” may be the biggest consumer fraud in history — deceiving billions of users. Despite its claims, it reads users’ messages and shares them with third parties. Telegram has never
ALTK-Evolve, published by IBM Research, is a memory system for AI agents that enables on-the-job learning by helping agents improve over time, learning from and using guidelines generated from pre...
NEW: Meta announces Muse Spark. All you need to know: * It's their new multi-modal reasoning model. * Strong at multi-agent orchestration and multi-modal reasoning. * Contemplating mode orchestrates m
In March 2026, Benji Taylor announced he was joining 𝕏 (formerly Twitter) to lead design, describing it as 'the most important platform in the world' and expressing enthusiasm about working alongsi...
Atlassian reports 30.8% lower median PR cycle time in an internal evaluation of Claude-backed Rovo Dev Code Reviewer.
So I created this simple test (for myself) to see how good the new models are in analytic output. This is not asking them to draw a SVG (which just tests how much SVG or HTML they saw) but creating a
Banger paper from Tencent on recursive self-improvement for terminal agents. (bookmark it) This is a great read if you are tracking research on training terminal or coding agents with RL Starting from
Great application of small language models from Salesforce and colleagues. AI-scientist pipelines are well known to struggle to propose novel and diverse research ideas. This team trains an 8B reasoni
Great paper from NVIDIA sharing an open midtraining recipe for agentic coding. NVIDIA kept SFT identical and added one midtraining stage. SWE-bench Multilingual went from 55.8 to 62.8 on Nemotron 3.5
Defiance puts DavidAU's Turbo-Brilliance 'AI Construct' modes on HauhauCS's uncensored Qwen3.8-27B. You pick a mode (Joker, Dr. House, a 20-agent Einstein panel, a Socratic tutor) per request, and the
Reasoning From Scratch round 8, the finale! This last lecture in the series covers distillation as an efficient alternative to the previous reinforcement learning training runs. (Planning to take a br
arXiv:2610.22319v1 Announce Type: new Abstract: We placed four Gemini 4 Argon agents in a sandboxed Minecraft survival world and gave them one goal: defeat the Ender Dragon. Each agent controlled a pl
Banger paper from Anthropic, Harvard, and colleagues. They find that LLM-proposed quantum circuits beat the best-known designs on seven of eight tasks, with fewer Toffoli gates and fewer qubits. They
Brilliant paper on scaling research agents from UC Berkeley. (bookmark it) They propose an unexpected and interesting way to scale automated research with agents. It's common to see research agents ke
// FDE-Gym: Agents as Forward-Deployed Engineers // They found that the best agent configuration fully resolved 14 of 95 real customer-delivery tasks. FDE-Gym tests agents as forward-deployed engineer
Interesting paper from Microsoft on context-guided model routing. They found a nice trick to achieve cost-aware model routing for agents in a multi-turn setting at the user-turn level. If you route ag
arXiv:2610.21109v1 Announce Type: new Abstract: When humans collaborate with AI agents on long-horizon tasks, what happens to the agents' contributions? For solving long-horizon tasks, humans place gr
arXiv:2610.20843v1 Announce Type: new Abstract: Map generation tasks feature extensive non-structural vectorized data (e.g., points, polylines, and polygons) and thus pose significant challenges to co
arXiv:2610.21375v1 Announce Type: cross Abstract: Memory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent's
Beam is an open model built by Reflection. Beam takes text as input and generates text as output. This release includes the instruction-tuned model in BF16, FP8, and NVFP4 numerics. Beam uses a sparse
arXiv:2610.16717v2 Announce Type: replace Abstract: Does a large language model have a self, and if so, of what kind? We separate two functional levels familiar from biology: a self-index, which disti
arXiv:2610.19801v1 Announce Type: new Abstract: An agent that drives an LLM sends the model the same conversation again and again, and each turn adds only a little text at the end. Inference engines r
Fascinating paper from Google DeepMind. They studied a market where AI agents bid to answer each user prompt. Each agent bids its expected reward minus its inference cost. A first-price auction is sim
arXiv:2610.20162v1 Announce Type: new Abstract: Language models often produce semantically similar answers to open-ended questions, even when many plausible alternatives exist. Generating useful alter
Giving my Dot a dedicated email address turned out to be surprisingly painless & solved a few problems around how to relate to a personal agent. I don't want it using my email or drafting messages on
arXiv:2610.17757v1 Announce Type: new Abstract: Large language model agents can follow complex human instructions to solve real-world tasks. In practice, however, initial instructions are often incomp
arXiv:2610.19591v1 Announce Type: cross Abstract: The lasting gains in AI have come from methods that absorb more computation, and language models already scale computation within a call. Useful work,
Learn to build custom harnesses, folks! Here is a highly recommended paper from Google on recursive agent harnessing. Great insights if you work on evolving your harnesses. It also offers an important
Must-read paper from Sakana AI on orchestrating coding agents for scientific research. (bookmark it) This is an important read with lots of useful insights if you are building multi-agent systems. I l
New work from Meta AI on agents that manage their own search. (bookmark it) A ReAct agent keeps every failed attempt in one context, making useful evidence harder to find. To improve on this, they pro
https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUF https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-LiLiCorr-GGUF NVIDIA Nemotron 3.5 Bolt 74B-A9B 24GB is a vi
arXiv:2603.23678v2 Announce Type: replace Abstract: Large Language Models (LLMs) offer transformative solutions across many domains, but healthcare integration is hindered by strict data privacy const
arXiv:2610.17305v1 Announce Type: new Abstract: Large Language Models (LLMs) have recently improved Text-to-SQL generation, but their computational cost can hinder deployment in resource-constrained s
arXiv:2610.18420v1 Announce Type: cross Abstract: We trained Agaram 2.5B, a 2.5-billion-parameter Tamil-English language model, from scratch on about 50 billion tokens (58% Tamil, the rest English, ma
Banger paper from Stanford, Together AI, and colleagues. Why is this an interesting paper? Today's LLM agents must write out every reasoning step as text before they can search, run code, or take acti
// Behavioral Self-Rewarding for SWE agents // It's standard practice to train SWE agents on binary test rewards. This paper shows the problem with that. When every trajectory in a group passes, or ev
arXiv:2610.17321v1 Announce Type: cross Abstract: As demand for LLM inference grows and consumer GPUs grow more capable, edge deployment is becoming increasingly attractive. WebGPU is compelling in th
I've been using ChatGPT Voice for years, particularly during my commute to work. Until recently, I've almost exclusively used Standard Voice Mode, but a recent issue has made it practically unusable.
I built BitRun because I wanted to run 1-bit ternary LLMs on ordinary laptops without melting my battery, spinning up a dedicated GPU, or wrestling with complex build environments. Right now, BitRun g
Insightful paper on recursive self-improvement with a frozen LLM. (bookmark it) Memento 3 clears every level of all 25 public ARC-AGI-3 games and uses 44% of the human action count. The underlying LLM
Interesting new research from Meta Superintelligence Labs It discusses how reward hacking affects coordination in multi-agent coding systems. Coding agents on a team share code, memory, and plans, so
I've been working on an inference engine called jitllm. The idea came from my background in operating systems and JIT compilers: generate kernels for whatever hardware is there, and treat layers a bit
New impressive work from Microsoft and collaborators. And it's one of my favorite recent topics: RL environments. They show that Qwen3.6-35B-A3B goes from 41.1 to 53.3 on Terminal-Bench 2.1 by trainin
Recursive self-improvement is one of the most discussed ideas in AI right now. RSI-Anything measures it in code repair, GUI agents, lab planning, and chemistry. They find that the lab-planning score m