Morning Brief: Monday, July 13

Sixty-five feeds. Two weeks. 4,372 items reduced to what follows. (what we track, how we crawl, subscribe)

Monday. The frontier-tension beat opens the day. Ray Myers amplifies a Zed-creator piece calling Anthropic out for blowing smoke — one of the sharpest first-party critiques of frontier-lab messaging since Simon Willison's summer of Fable posts. Cory Doctorow files a companion argument on the AI-companies-vs-their-customers inversion. Berkshire Hathaway's $397B short of the US market surfaces in the HN top-ten — read against Interconnects' Sunday 6 months to live for open models post, both frame the AI capex cycle as approaching an inflection point. Nature Machine Intelligence lands a fresh package (three items): The brain is a diverse place, why not computing?, the ergoCub humanoid embodiment paper, and A manifesto for Sustainability Robotics. Interpretability push visible on Pinboard — Anthropic's Circuit Tracing methods paper resurfaces alongside a Jacobian Lens share on Qwen3.6-27B via Neuronpedia. Terence Tao files a small tools post (A random variable visualizer). arXiv cs.AI's Monday drop is 158 items; the CLI-agent failure-mode agenda hardens with Failure as a Process and Long-Horizon-Terminal-Bench. Sunday alignment tail still active: Alignment Forum's Independent alignment and From wantons to moral agents.

Top (5-7 min)

Zed Creator Calls Spade a Spade, Anthropic Blows Smoke: HN, 2026-07-13. Ray Myers on the Zed editor creator's public pushback against Anthropic's rhetoric. Read alongside Kenneth Payne's Friday Taste the difference and the ongoing Fable bump thread from Simon Willison — three independent voices reading the frontier-lab narrative critically in one week.
Pluralistic: Why aren't AI companies competing directly with their customers?: Pluralistic, 2026-07-13. Doctorow's Monday essay on the AI-vs-customer inversion. Companion frame to the Zed piece above and to Interconnects' 6 months to live for open models.
Berkshire's $397B Bet Against an Overheated Market: HN, 2026-07-13. Berkshire files a market-short at scale. HN commenters read the AI capex cycle as the visible target. Pair with the Saturday circular financing of the GPU boom Nvidia/CoreWeave/Nebius piece from IO Fund still on the Sunday reading list.
6 months to live for open models: Interconnects, 2026-07-12. Nathan Lambert's Sunday piece — the highest-signal frontier-lab read of the weekend. Read with the Zed/Doctorow/Berkshire trio above as a coherent the party is peaking frame.
The brain is a diverse place, why not computing?: Nature Machine Intelligence, 2026-07-13. Monday NMI cover-package essay on brain-inspired computing diversity. Companion papers below: an ergoCub humanoid embodiment paper and a manifesto for sustainability robotics.
Failure as a Process: An Anatomy of CLI Coding Agent Trajectories: arXiv cs.AI, 2026-07-13. Direct follow-through on Saturday's 85% say code review is the new bottleneck and Friday's Every IC engineer is really a front line manager now: the failure-mode literature on CLI coding agents is now a distinct subfield.
Circuit Tracing: Revealing Computational Graphs in Language Models: Pinboard, 2026-07-13. Anthropic's Circuit Tracing methods paper (via transformer-circuits) bookmarked today. Read with Friday's MIT TR Anthropic found a hidden space where Claude puzzles over concepts and today's Jacobian Lens Qwen3.6-27B share on Neuronpedia.

Themes this week

Frontier-lab critique from independent voices: Ray Myers: Zed creator calls spade a spade, Anthropic blows smoke (today), Pluralistic: Why aren't AI companies competing directly with their customers? (today), Interconnects: 6 months to live for open models (Sun), Kenneth Payne: Taste the difference (Fri), Simon Willison: Fable gets another bump (Sun).
AI capex cycle stress signals: HN: Berkshire's $397B bet against an overheated market (today), IO Fund: Nvidia/CoreWeave/Nebius circular financing of the GPU boom (Sat), TNS: OpenAI, SpaceXAI, Meta race to the bottom on price (Sat).
CLI coding-agent failure modes as literature: arXiv: Failure as a Process — Anatomy of CLI Coding Agent Trajectories (today), arXiv: Long-Horizon-Terminal-Bench (today), arXiv: The Patchwork Problem in LLM-Generated Code (today), arXiv: SCATE — Learning to Supervise Coding Agents for Cost-Effective Test Generation (today), arXiv: Inside the Skill Market — From Software Engineering Activities to Reusable Agent Skills (today), TNS: 85% say code review is the new bottleneck (Sat).
Interpretability tools going public: transformer-circuits: Circuit Tracing methods paper (via Pinboard, today), Neuronpedia: Jacobian Lens on Qwen3.6-27B (via Pinboard, today, also seen Sun), MIT TR: Anthropic found a hidden space where Claude puzzles over concepts (Fri).
Nature Machine Intelligence cover package: NMI: The brain is a diverse place, why not computing? (today), NMI: Towards shared embodied intelligence in humanoid robots — ergoCub (today), NMI: A manifesto for Sustainability Robotics (today), NMI: Four questions for AI-ready biological data (Fri).
Alignment theory tail: Alignment Forum: Independent alignment of language models (Sun), Alignment Forum: From wantons to moral agents (Sun), Alignment Forum: The current bottleneck is political will, not research (Sat).
Mathematicians on tools: Terence Tao: A random variable visualizer (today), Terence Tao: Old and new apps via modern coding agents (Sat), Andrej Bauer: Making AI smarter with AI (Sat).

Scan (15 min)

Monday signal (short)
- Zed Creator Calls Spade a Spade, Anthropic Blows Smoke, HN, 07-13
- Pluralistic: Why aren't AI companies competing directly with their customers?, Pluralistic, 07-13
- Berkshire's $397B Bet Against an Overheated Market, HN, 07-13
- America May Soon Be Facing Its Largest Labor Shortage in Its History, Slashdot, 07-13
- Japan's Space Agency Conducts First Test Flight For Experimental Reusable Rocket, Slashdot, 07-13
- Semi-Trailer Trucks Test Converting Into Plug-In Hybrids, Slashdot, 07-13
- A random variable visualizer, Terence Tao, 07-13
- JetZero's changed Z4 BWB, Part 3: the business plan, Leeham News, 07-13
HN Monday tabs
- Quadrupling code performance with a "useless" if, HN, 07-13
- Backtrack-Free Cursive, HN, 07-13
- The Graph That Should Be Front-Page News, HN, 07-13
- Ask HN: Add flag for AI-generated articles, HN, 07-13
- Beavis Ultrasound PnP ISA Sound Card Replica, HN, 07-13
- Interrail: 6,379Km and 13 Countries over 7 weeks, HN, 07-13
- Sam Neill has died, HN, 07-13
Nature Machine Intelligence — Monday cover package
- The brain is a diverse place, why not computing?, NMI, 07-13
- Towards shared embodied intelligence in humanoid robots (ergoCub), NMI, 07-13
- A manifesto for Sustainability Robotics, NMI, 07-13
Sunday alignment carryover (worth reading Monday)
- Independent alignment of language models, Alignment Forum, 07-12
- From wantons to moral agents, Alignment Forum, 07-12
- 6 months to live for open models, Interconnects, 07-12
- Directly Responsible Individuals (DRI), Simon Willison, 07-12
- sqlite-utils 4.1.1, Simon Willison, 07-12
- shot-scraper 1.11, Simon Willison, 07-12
- Fable gets another bump, Simon Willison, 07-12
- Kernel prepatch 7.2-rc3, LWN, 07-12
- Estimating Vocabulary Size with a Simple Bayesian Model, Planet Clojure, 07-12
- KiCad helpers and caliper improvements, Planet Clojure, 07-12
arXiv cs.AI Monday drop (158 items — CLI-agent failure-mode core)
- Failure as a Process: An Anatomy of CLI Coding Agent Trajectories, arXiv, 07-13
- Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks, arXiv, 07-13
- The Patchwork Problem in LLM-Generated Code, arXiv, 07-13
- SCATE: Learning to Supervise Coding Agents for Cost-Effective Test Generation, arXiv, 07-13
- Inside the Skill Market: From Software Engineering Activities to Reusable Agent Skills, arXiv, 07-13
- Scoped Verification for Reliable Long-Horizon Agentic Context Evolution under Distribution Shift, arXiv, 07-13
- Toward Auditable AI Scientists: A Hypothesis Evolution Protocol for LLM Agents, arXiv, 07-13
- OpenProver: Agentic and Interactive Theorem Proving with Lean 4, arXiv, 07-13
- ProofCouncil: An LLM Agent for Solving Open Mathematical Problems, arXiv, 07-13
- Shared Selective Persistent Memory for Agentic LLM Systems, arXiv, 07-13
- Ceci n'est pas une pipe: AI systems as semantic abstractions, arXiv, 07-13
- Multimodal Reward Hacking in Reinforcement Learning, arXiv, 07-13
- ConceptSMILE: Auditing the Trustworthiness of Concept-Based Explainable AI, arXiv, 07-13
- GATS: Graph-Augmented Tree Search with Layered World Models for Efficient Agent Planning, arXiv, 07-13
Interpretability tools (Pinboard bookmarks today)
- Circuit Tracing: Revealing Computational Graphs in Language Models, Pinboard, 07-13
- Jacobian Lens — Qwen3.6-27B on Neuronpedia, Pinboard, 07-13
Hackaday build log
- Newton's Cradle Isn't Really Perpetual, Hackaday, 07-13
- CGI Motion Capture with only a Camera, Hackaday, 07-13
- Musing on AI from 1964, Hackaday, 07-13
RTL-SDR (three-item Monday drop)
- Airspy WebSpy: A High Performance Web UI Client for Airspy SDRs, RTL-SDR, 07-13
- Exploring the Art and Science of Spectral Painting with SDR, RTL-SDR, 07-13
- A Paper on Clock Distribution for Phase Coherent Operation of RTL-SDR Receivers, RTL-SDR, 07-13

Tail

Special Seminar with Nachum Ulanovsky: Neural codes for natural behaviors in bats, MIT calendar, 07-13
Engineering Business Certificate — Galante Info Session, Northeastern events, 07-13

Feed silences (diagnostic)

arxiv-cs-ai: 158 items today (Monday drop). Saturday was 216 items; Sunday cliff was 0. The CLI-agent failure-mode subgenre is the distinct signal in the drop.
Anthropic first-party, OpenAI, Simon Willison, Apple ML Research, Google Research, ClickHouse, EFF Deeplinks, Schneier, GitHub Blog, Interconnects, Kenneth Payne, Martin Fowler, Microsoft Research, Cloudflare, Charity Majors, Databricks, Hugging Face Blog, MIT Technology Review, Alignment Forum, Quanta Magazine, Grafana Labs, Vercel, The New Stack, 404 Media, Latent Space, TechCrunch, Planet Clojure, LWN, Andrej Bauer: no primary posts today (Monday morning silence — the first-party publishing channels typically wake up mid-morning US-East).
Netflix Tech Blog, James Bornholt: persistent DNS/TLS errors continue (unchanged from Sun).
WBUR: no Endless Thread today (weekly cadence).
bitsavers (6 feeds): connected, 0 items (sparse archive, expected).
claude-code-releases: no release today.

Build provenance

build: 2026-07-13 | crawler-sha: 5fe7ab8 (Walsh-Research/1.2, compliance v1.3) | feeds: 65 active (78 configured, incl. 12 corp-eng, 6 bitsavers, 5 generated) | items-considered: 4372 (14d, incl. 2555 arXiv) | warehouse: 24619 items | published: 7 | note: Monday. Frontier-lab critique surfaces from three independent voices — Ray Myers/Zed creator on Anthropic, Doctorow on AI-vs-customer inversion, Interconnects (Sun) on open models' 6-month horizon. Berkshire's $397B market-short reads to HN as AI capex bet. Nature Machine Intelligence Monday cover-package: brain-inspired computing, ergoCub humanoid, sustainability robotics manifesto. Interpretability tools on Pinboard (Circuit Tracing, Jacobian Lens). arXiv cs.AI Monday drop 158 items — CLI-agent failure-mode subgenre hardens (Failure as a Process, Long-Horizon-Terminal-Bench, Patchwork Problem, SCATE, Skill Market).