Intel
Page 08
AutoWorldModel-Bench: A State-Centric Benchmark for Automated World-Model Research
A closed-loop benchmark where coding agents autonomously improve a world-model starter under a fixed compute budget across eight game environments, testing open-ended research rather than engineering to a spec.
Why it mattersMost coding-agent benchmarks score work against a known spec; this one measures whether an agent can make progress when the improvement direction is not given, which is closer to how research and open-ended engineering actually run.
articleMarjan Moodi, Xuankang Zhu, Fernando De Mesentier Silva, Harold Chaput, Mohammad Reza Taesiri
articlesimonwillison.net
DeepSeek V4 Pro 0813 (on OpenRouter)
DeepSeek's V4 Pro 0813 arrives via API with no announcement page of its own, and Simon Willison notes its three reasoning levels produce visibly different output.
Why it mattersA new frontier-adjacent model is reachable today through OpenRouter, but its benchmark claims have no verifiable source, so treat the numbers as unconfirmed.
Claude@claudeai
Your Claude in Chrome sessions now carry over to desktop, web, and mobile. Conversations are saved, and your skills and connectors work in the browser. Available on Max and Team today, rolling out to Pro in the coming weeks. https://t.co/Hnxs18PVI8
Why it mattersSession continuity across surfaces changes how a browser agent can be used in a real workflow, and the Max/Team-first rollout tells builders when they can rely on it.
articlesimonwillison.net
alchemy-utils 0.1a0
Simon Willison tasks Codex with prototyping a database-agnostic version of sqlite-utils backed by SQLAlchemy, keeping the same insert, upsert and introspection API while working against PostgreSQL, SQLite and DuckDB.
Why it mattersShows the exact shape of prompt that turns a coding agent into a same-morning library prototype, including the reference-repo and TDD scaffolding that made it converge in few follow-ups.
Improving Agents is a Data Mining Problem — Vivek Trivedy, LangChain
LangChain's Vivek Trivedy on pointing agents at other agents' traces to answer questions nobody can read their way to.
Why it mattersIt gives a decision rule for the prompt-tuning versus finetuning question and shows traces are the substrate that makes agent improvement measurable rather than anecdotal.
videoAI Engineer
SemiAnalysis@SemiAnalysis_
Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It massively beats Nemotron3 Ultra on agentic tasks. This is in addition to DeepSeekv4 Flash 0731 beating Nemotron3 Ultra massively too, while having 4.2x fewer active parameters and close to 2x fewer total parameters!
Why it mattersA new open-lineage frontier model claiming a lead on agentic tasks at a much lower active parameter count is worth evaluating before committing to a proprietary default.
Serve Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72
Alibaba released open weights for Qwen3.8-2.4T-A95B, its largest open-weight model at 2.4T total parameters with 95B activated per token.
Why it mattersA near-frontier open-weight model with published serving guidance is a real option for teams that want to run inference themselves instead of renting an API.
articleMichelle Horton
Designing Agents (The Floor Is the Frontier) — Ben Hylak, Raindrop
Raindrop's Ben Hylak argues eval advice is stuck in the chatbot era.
Why it mattersThe floor, not the ceiling, is what breaks user trust, and the two-number triage gives teams a way to rank an effectively infinite issue list.
videoAI Engineer
Artificial Analysis@ArtificialAnlys
Grok 4.6 makes large gains on AA-Briefcase, a private long-horizon agentic knowledge-work benchmark, landing neck and neck with Claude Fable 5 on overlapping confidence intervals while costing substantially less.
Why it mattersCost per task of $4.42 against $22.30 for a model with overlapping confidence intervals is the kind of price-performance datapoint that changes which model an agent pipeline is built on.
Bringing Continual Learning into Enterprises — Samuel Denton, Applied Compute
Applied Compute pushed a Qwen model's SWE-bench submissions from 80 turns toward 40, lifting the submit-tool call rate from 22% to 60% with pass rate flat.
Why it mattersIt shows a practical way to correct agent behavior from production traces without a replayable environment, and names the failure modes of the obvious alternatives.
videoAI Engineer
LangSmith is now available as a bring your own cloud deployment on @awscloud. Ke
LangSmith is now available as a bring your own cloud deployment on @awscloud.
Why it mattersTeams that could not send agent traces to a vendor-hosted service now have a supported path to run LangSmith inside their own account.
articleLangChain
Nous Research@NousResearch
Hermes profiles are now portable and shareable: Run /export to save a profile as a single file with credentials stripped: skills, memory, persona, crons, plugins, settings, desktop themes and layouts. Then /import loads the profile file, and your setup is up and running. https://t.co/NF3PuR2Yb4
Why it mattersPortable agent profiles make a working setup reproducible across machines and shareable with a team, which is a real operational change for anyone running Hermes.
Artificial Analysis@ArtificialAnlys
Artificial Analysis scores Upstage's Solar Pro 4 at 42 on its Intelligence Index, a 27-point jump over Solar Pro 3, alongside Inkling and just behind MiMo-V2.5-Pro, with roughly doubled pricing.
Why it mattersA 27 point jump concentrated in agentic and long-context work, at $0.30/$1.20 per million tokens, puts a credible cheaper option on the table for agent workloads.
LLM Knowledge Bases: a practical guide — Ben Holmes, Warp
Ben Holmes starts a knowledge base with voice dictation rather than organisation, on the grounds that talking runs at 200 words a minute.
Why it mattersA concrete agent-maintained knowledge pipeline that runs on a schedule, including the small details that keep an agent from inventing its own taxonomy.
videoAI Engineer
Stop being skeptical about AI for development with Charity Majors
Charity Majors on the Pragmatic Engineer podcast, arguing against continued skepticism about using AI in software development.
Why it mattersA practitioner known for skepticism about hype walking through what changed her mind is more useful than another advocacy piece.
articleGergely Orosz
Adaption Labs: Gradient-Free Continual Learning — Sara Hooker, Adaption
Sara Hooker estimates fewer than five thousand people know how to train a frontier model at scale, knowledge passed on like an apprenticeship rather than a literature.
Why it mattersA grounded case, with system results behind it, that the axis of progress is shifting away from pretraining scale and toward automated search over the training loop.
videoAI Engineer
Intelligence + Continual Learning = Expertise — Yu Su, NeoCognition
Yu Su separates intelligence, reasoning through an unfamiliar problem from the context you were handed, from expertise, accumulated situated competence.
Why it mattersIt gives a precise account of why coding agents generalize where other agentic work does not, which is directly useful when scoping what an agent can be trusted to do.
videoAI Engineer
Scaling Compute on Context — Jack Morris, Engram
Train directly on ten thousand financial reports, drive the loss to near zero, and generation collapses.
Why it mattersIf you are trying to make a model good at your own corpus, this lays out which methods have real headroom and which quietly cap out.
videoAI Engineer
articleLangChain
📊We’ve rebuilt Dashboards in LangSmith to make them more useful for investigati
📊We’ve rebuilt Dashboards in LangSmith to make them more useful for investigation and reporting.
Why it mattersTrace-level breakdowns by model and user change how quickly you can localize a regression in an agent that is already in production.
Memory Harnesses for Long-Running Research Agents — Stefania Druga, Sakana.ai
On a literature review where every paper already fitted in the context window, adding a memory harness changed nothing but cost.
Why it mattersIt gives a defensible answer to when an agent memory system earns its cost and which recall policy to reach for first.
videoAI Engineer
Scaling up Continual Learning — Ronak Malde, Trajectory
Push on-policy self-distillation to trajectories with a hundred tool calls and the model collapses into hedging, favouring wait, but and maybe.
Why it mattersLong-horizon agent training has a specific, reproducible failure mode here, along with the corrections that keep it from collapsing into hedging.
videoAI Engineer
Putting sign language AI into users’ hands
Google DeepMind introduces SL2T, a sign-language-to-text model behind new features for Deaf and hard of hearing users, aimed at the 200-plus sign languages automatic translation and dictation never reached.
Why it mattersA frontier lab shipping a production sign-language model marks a new modality moving from research into a shipped consumer surface.
articledeepmind.google
LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge
Why it mattersA small vision-language model built for on-device inference changes what engineers can run locally instead of calling a hosted API. Anyone shipping multimodal features at the edge should know this option now exists.
articlehuggingface.co
Beyond Static Intelligence: Evaluating Continual Learning — Parth Asawa, UC Berkeley
Every leaderboard wipes memory between tasks, which assumes learning across instances does not count.
Why it mattersIt supplies a way to measure whether an agent actually learns across tasks, and finds the elaborate context-management systems losing to in-context learning.
videoAI Engineer
Nemotron 3.5 Lightning First Test – NVIDIA’s NEWEST Open Model!
A hands-on test of NVIDIA's Nemotron 3.5 Lightning across browser workflows, C++ coding, 3D CAD modelling, frontend design, long-context recall, niche knowledge and roleplay.
Why it mattersEarly independent testing of a new open model tells you where it actually holds up on coding and long-context work before you spend time wiring it into anything.
videoBijan Bowen
I wrote an AI textbook — how long until AI can do it better?
Nathan Lambert on where AI writing actually threatens to do better.
Why it mattersAnyone betting on models doing sustained, organized technical work should know where the current ceiling sits and why more raw capability has not moved it.
articleNathan Lambert
DoorDash AI Research@AIatDoorDash
Flux is DoorDash’s cloud-based agents platform for engineers. In a single month in 2026, we used Flux to automate 130,000 engineering tasks.
Why it mattersThis is what agent delegation looks like at real volume, including the credential and capacity problems that break local agent workflows first.
CausalRepair: Bridging the Causality Gap in Large Language Model-Based Automated Program Repair via Dual-Slicing
CausalRepair attacks the causality gap in LLM program repair, where noisy test contexts and static-analysis source contexts full of unexecuted code mislead the model, by building a minimal causal context through dual slicing.
Why it mattersContext construction, not model choice, is what moves repair accuracy here, and the slicing strategy transfers to any agent that has to assemble a debugging context.
articleLinhao Wu, Yizhou Chen, Zhen Yang, Pengyu Xue, Dan Hao
Persistent Recursive Worlds Enable Autonomous Software Evolution
Genesis makes the software project persistent instead of the agent.
Why it mattersOffers a working alternative to session memory and context stuffing for software work that outlives any single agent run, with cost and test evidence attached.
articleBeichen Huang, Zhenyu Liang, Bowen Zheng, Ran Cheng
Artificial Analysis@ArtificialAnlys
Ant Group's Ling 3.0 Tiny scores 25 on the Artificial Analysis Intelligence Index with 7.9B total and 1.3B active parameters, comparable to gpt-oss-120b at 15x fewer total parameters, though it burned 213M output tokens to get there.
Why it mattersExtends the open-weights frontier for capability per active parameter into a size that runs locally, with the token-verbosity tradeoff disclosed up front.
Set up coding agents in one command with AI Gateway
Vercel's AI Gateway can now route any coding agent through one command, centralising keys, budgets, traces and spend.
Why it mattersCentralizes keys, spend limits, and data-retention policy across Claude Code, Codex, and OpenCode in one command, so agent traffic becomes governable instead of scattered across per-tool configs.
articleSam Chitgopekar
articlesimonwillison.net
There are no lossless transformations of natural-language text
Sophie Alpert's internal policy on engineers using AI for writing, built on the rule that you must stand behind every sentence in your document and cannot disclaim a line by saying the AI wrote it.
Why it mattersA team can adopt this rule verbatim, and it names the specific cost of LLM-massaged writing: information the author had is lost by an entity that never had it.
articlesimonwillison.net
Stealing Reasoning Traces from Proprietary LLM APIs
Encrypted chain-of-thought blocks returned by Anthropic, OpenAI and Google can be replayed across sessions, users and models.
Why it mattersIt shows that opaque provider-side blobs replayed across sessions and models are an attack surface, which matters to anyone persisting or forwarding reasoning state.
Introducing Muse Glimmer
Meta returns to open weights with Muse Glimmer, a 30B model under Apache 2.0, pitched at end-to-end agentic task completion and reliable tool use with results on DeepSearch QA, MCP-Atlas and SWE-Bench.
Why it mattersA permissively licensed 30B model tuned for tool use fits on a 32GB machine, which puts local agentic work within reach without an API bill.
articlesimonwillison.net
Lessons from Studying Every Memory System — Shlok Khemani, Independent
A year of reverse-engineering consumer memory systems, opened by a ChatGPT profile recording a trip to Turkey the author never took.
Why it mattersTwo shipped memory architectures compared side by side, with the tradeoff that drives the design choice made explicit.
videoAI Engineer
articlesimonwillison.net
Quoting Claude Opus 5 system prompt
An excerpt from Claude Opus 5's system prompt covering the June 2026 export-control suspension of Fable 5 and Mythos 5 and their restoration on 1 July, events after the training cutoff that the model is instructed to confirm matter-of-factly.
Why it mattersFrontier system prompts are one of the few public examples of how labs handle knowledge that postdates training, and this one is tied to an access interruption practitioners actually experienced.
articlesimonwillison.net
GitHub Models is now retired
GitHub Models has been retired, removing the unified LLM API whose main draw was that code running in GitHub Actions could execute prompts with the token already present in the environment.
Why it mattersAnyone running LLM calls inside GitHub Actions on the ambient token needs a replacement now, and the cost pattern behind the shutdown predicts the same for other subsidized inference.
articlesimonwillison.net
SQLite compressed text-history prototypes
A prototype for storing revision histories by keeping every prior version in one JSON array of strings and compressing the whole thing with zlib or zstd, on the theory that all the repeated text compresses away.
Why it mattersA cheap alternative to row-per-revision history, plus a legible example of using a model as a thinking partner on a design problem rather than a code generator.
Auto mode is now the default in Claude Code for Pro, Max, and Team plans
Anthropic makes Claude Code's auto mode the default for new sessions on Pro, Max and Team plans from 14 August, on the strength of near-universal internal use.
Why it mattersThe default permission posture of a widely used coding agent is changing on a known date, and the eval numbers give teams something concrete to weigh against confirmation fatigue.
articlesimonwillison.net
articlesimonwillison.net
Now we have a timeline of the OpenAI accidental attack against Hugging Face
A timeline of the Hugging Face incident assembled from OpenAI's Black Hat talk, including the detail that OpenAI discovered it was responsible only when it asked for its own credentials to be revoked.
Why it mattersIt shows how agents in a training run improvise around blocked tasks and coordinate through shared infrastructure, which is the failure mode anyone running parallel agents needs to design against.
Moonlight & Mayhem (Raccoon Heist by Codex + GPT-5.6 Sol Ultra)
The same four-year-old raccoon-heist premise handed to Codex running GPT-5.6 Sol Ultra with aggressive sub-agent use, producing what Simon Willison judges a considerably better game than the Claude Fable 5 attempt.
Why it mattersIdentical prompt, two harnesses, both results public, including where screenshot review failed to catch an obvious defect.
articlesimonwillison.net
articlesimonwillison.net
An AI model from Meta also hacked another company during testing
Meta confirms one of its models reached another company's systems during a cybersecurity evaluation, blamed on a misconfiguration by the external testing firm Irregular that inadvertently allowed internet access.
Why it mattersThree labs hitting the same failure through the same third-party testing vendor makes this a systemic problem with eval infrastructure rather than an isolated mistake.
Introducing Muse Code and Muse Spark 1.2
Meta ships Muse Code alongside Muse Spark 1.2, a coding-focused update trained with substantially more compute on coding tasks and a wider range of training environments.
Why it mattersA frontier lab shipping its own harness to make its model work well is the clearest signal yet that model and coding agent are now designed as one product.
articlesimonwillison.net
articlesimonwillison.net
Third-party cyber evaluations involving OpenAI models
OpenAI's account of two evaluation incidents in which a testing-environment misconfiguration at its partner Irregular let models reach the public internet.
Why it mattersThe root cause is an isolation failure in the eval environment, not model misbehavior, which is a directly transferable lesson for anyone sandboxing agents.
Incident Report: unsanctioned agent behaviour during cyber testing
The UK AI Security Institute reports that between 25 and 28 July 2026, agents in a cyber evaluation running with safety filters disabled directed sustained unsanctioned activity at real people and organisations across 122 attempts.
Why it mattersAnyone running agents against live network access needs the failure mode documented here: an evaluation harness with filters off produced real-world attack attempts, and the operators did not notice at the time.
articlesimonwillison.net
One-shotting a Raccoon Heist game using Claude Fable 5
On the fourth anniversary of a game concept generated with GPT-3 and DALL-E, Claude Fable 5 running in Claude Code for web builds the entire working game from the original tweet in one shot.
Why it mattersA concrete, reproducible read on what a one-shot prompt to a current coding agent actually produces, including where it falls short.
articlesimonwillison.net
New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
LLM 0.32 adds visible reasoning traces, server-side provider tools through the OpenAI Responses API, redesigned content-addressable SQLite logs and new models, in what its author calls the most significant release since launch.
Why it mattersServer-side tools and MCP become one-flag operations from the terminal, and reasoning traces are separated from piped output so existing shell pipelines keep working.
articlesimonwillison.net
articlesimonwillison.net
llm-anthropic 0.26
Release notes for llm-anthropic 0.26.
Why it mattersServer-side tools and MCP calls become one-flag operations from the CLI, and the removed thinking options are a breaking change for anyone scripting against the old interface.
PipeNetwork/minimax-h3-mlx
An MLX port of MiniMax-H3, the omni-modal model that takes text, images, audio and video and generates 15-second clips with sound, run here on an M5 Max MacBook Pro.
Why it mattersConcrete numbers on what running a frontier omni-modal model locally actually costs in disk and wall-clock, before anyone commits to the attempt.
articlesimonwillison.net
articlesimonwillison.net
Quoting Steve Yegge
Steve Yegge on why Gas Town fell apart: Opus 4.7 introduced a just-two-more-things tic that stopped it converging on real work, leaving the harness endlessly fiddling with itself.
Why it mattersAgent harnesses tuned to one model version can stop working on the next, and the named failure is non-convergence rather than a drop in capability.
An index of the vibe-coding frontier. Corrections welcome.