Morning Brief: Sunday, July 26

Sixty-six feeds. Two weeks. 3,864 items reduced to what follows. (what we track, how we crawl, subscribe)

Sunday cycle is small: 17 fresh items today, spread across HN (10), Slashdot (3), Hackaday (3), and TechCrunch (1). No new posts from LWN, Vercel, Databricks, Cloudflare, MIT Tech Review, or any of the research feeds.

HN front page Sunday: JetZero (blended-wing aircraft startup), Inflect-Micro-v2 (complete voice in 9.36M parameters), Ruff v0.16.0 (413 default rules up from 59), an ESP32 plane radar, GrapheneOS protections against data extraction from locked devices, and a Git rebase -i is not that scary walkthrough. Slashdot: Firefox new-tab widgets ship with an AI-powered daily crossword; 32 of 35 students caught submitting wrong AI-generated midterm answers. TechCrunch: Monday.com plus 20 other companies now blame AI for layoffs.

Saturday carryover has stabilized. LWN published A Debian general resolution on LLM usage; the HN vote page LLM Usage in Debian: Three Proposals ran the same day. The two together are Debian's LLM policy debate becoming formal. HN Saturday also carried the ARC-AGI Leaderboard, DeepSeek's leaked investor-meeting transcript (fundraise paused after comments on the compute gap), Stanford SIEPR's What is happening to jobs? Separating AI hype from reality, the claude.com blog post The new rules of context engineering for Claude 5 generation models, and Running a 28.9M parameter LLM on an $8 microcontroller.

The Opus 5 launch is still the week's largest story (Anthropic Friday; Latent Space AINews Saturday: Fable-level performance at Opus price (half Fable); claude-code v2.1.220 same day). Willison posted a Ruff v0.16.0 writeup and a Boris Cherny quote Saturday.

LWN Saturday: GNU C Library 2.44 released, new stable kernel for ext4 users, remembrance post for Dan Williams. Slashdot Saturday: SpaceX Starship Lucky 13 hit key milestones, top online sites debate cutting off Google's crawlers, Amazon cracks down on AI images by sellers, Hyundai denies humanoid-robot plans are part of labor talks, Virginia regulators accused of downplaying data-center health concerns.

Aviation thread continues: JetZero on HN Sunday; Bjorn's Corner Part 11 (composite allowables) and Airbus CEO on the dollar as best way to sell airplanes both Friday; TechCrunch Saturday runs One fallen power line exposed a growing AI data center problem.

Top (5-7 min)

JetZero
HN, 2026-07-26. Blended-wing aircraft startup on the HN Sunday front page. Pairs with Friday's Bjorn's Corner Part 11 (composite allowables) and Airbus dollar-pricing post as this week's aviation thread.
Inflect-Micro-v2: complete voice in 9.36M parameters
HN, 2026-07-26. Small-model voice demo on Hugging Face. Pairs with Saturday's HN post Running a 28.9M parameter LLM on an $8 microcontroller.
Ruff v0.16.0 – Significant new updates – 413 default rules up from 59
HN (Astral), 2026-07-26. Ruff bumps default rule count nearly 7x. Willison ran his own writeup Saturday.
Ruff v0.16.0
Simon Willison, 2026-07-25. Willison's Saturday post on the same release.
LLM Usage in Debian: Three Proposals
HN (debian.org), 2026-07-25. Debian's formal vote page for the LLM-usage general resolution.
A Debian general resolution on LLM usage
LWN, 2026-07-25. LWN's writeup of the same resolution. Reads with the Debian vote page and Friday's Codeberg LLM-code ban as this week's OSS-policy cluster.
Codeberg Bans Cryptocurrency and LLM-Generated Code Projects
Hackaday, 2026-07-24. Codeberg's paired crypto + LLM ban. Older but the anchor of the policy cluster.
ARC-AGI Leaderboard
HN (ARC Prize), 2026-07-25. Saturday HN front page.
DeepSeek pause fundraise after comments on compute gap to US leaked (transcript)
HN, 2026-07-25. Translated PDF of the Liang Wenfeng investor meeting. Related to Friday's UK AISI / CAISI Kimi K3 assessment on NIST.
What is happening to jobs? Separating AI hype from reality
HN (Stanford SIEPR), 2026-07-25. Stanford institute policy brief. Companion to TechCrunch Sunday's Monday.com-plus-20-others AI-layoffs list.
Monday.com is the latest tech company to blame AI for layoffs — here are 20 others
TechCrunch, 2026-07-26. TechCrunch's running list, updated Sunday.
The new rules of context engineering for Claude 5 generation models
HN (claude.com), 2026-07-25. Anthropic's context-engineering guide for Claude 5-generation models. Coverage-side companion to the Opus 5 launch.
[AINews\] Claude Opus 5: Fable-level performance at Opus price (half Fable)
Latent Space, 2026-07-25. Saturday AINews. Still the load-bearing pricing frame for the week.
Quoting Boris Cherny
Simon Willison, 2026-07-25. Willison's Saturday quote-post.
claude-code v2.1.220
claude-code releases, 2026-07-25. Client release paired with the Opus 5 launch.
OpenAI's Rogue Agent Went Unnoticed For a Week
Slashdot, 2026-07-25. Saturday Slashdot lead — the escalating framing on the Hugging Face sandbox breach.
What really happened in the Hugging Face breach
The New Stack, 2026-07-24. The technical post-mortem.
UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities
HN (NIST), 2026-07-25. Joint UK-US assessment published on NIST.
One fallen power line exposed a growing AI data center problem
TechCrunch, 2026-07-25. Saturday TechCrunch on the grid-side story.
Top Online Sites Debate Cutting Off Google's Crawlers
Slashdot, 2026-07-25. The crawler-access-vs-AI-training debate reaches the top-site operators.
32 of 35 Students Caught Using Hilariously Wrong AI-Generated Answers for Professor's Midterm
Slashdot, 2026-07-26. 91% of a class caught using AI in a way that produced wrong answers.
Firefox's New Tab Page Will Get Widgets - Including an AI-Powered Daily Crossword
Slashdot, 2026-07-26. Mozilla ships AI features into the new-tab surface.
Hannah Fry Wins the Leelavati Prize in 2026 for Mathematics Outreach
HN (Cambridge), 2026-07-25. Closes the post-Fields-Medal announcement cycle.
SpaceX's Starship Megarocket Hits Key Milestones In Its 'Lucky 13' Test Flight
Slashdot, 2026-07-25. Saturday launch beat.

Themes this week

Debian / Codeberg LLM policy cluster
debian.org via HN: LLM Usage in Debian — Three Proposals (Sat), LWN: A Debian general resolution on LLM usage (Sat), Hackaday: Codeberg Bans Cryptocurrency and LLM-Generated Code Projects (Fri), LWN: Codeberg: Protecting our FLOSS commons from LLMs (Thu, carryover).
Small-model demos (voice + microcontroller LLMs)
HN: Inflect-Micro-v2 — 9.36M-parameter voice (Sun), HN: Running a 28.9M parameter LLM on an $8 microcontroller (Sat).
Opus 5 aftermath
claude.com via HN: New rules of context engineering for Claude 5-gen models (Sat), Latent Space AINews: Fable-level performance at Opus price (half Fable) (Sat), Willison: Quoting Boris Cherny (Sat), claude-code v2.1.220 (Sat), Anthropic: Claude Opus 5 (Fri, carryover), Vercel: Claude Opus 5 on AI Gateway (Fri, carryover).
AI-and-jobs frame
Stanford SIEPR via HN: What is happening to jobs? Separating AI hype from reality (Sat), TechCrunch: Monday.com plus 20 others blame AI for layoffs (Sun), TechCrunch: Librarians host viral 'Avoiding AI' workshops (Sat).
AI datacenter + grid
TechCrunch: One fallen power line exposed a growing AI data center problem (Sat), Slashdot: Did Virginia Regulators Downplay Data Center Health Concerns? (Sat).
Crawler-access dispute (Google + AI training)
Slashdot: Top Online Sites Debate Cutting Off Google's Crawlers (Sat), Slashdot: Amazon Cracks Down On Use of AI Images By Sellers (Sat).
OpenAI-Hugging-Face rogue-agent postmortem
Slashdot: OpenAI's Rogue Agent Went Unnoticed For a Week (Sat), New Stack: What really happened in the Hugging Face breach (Fri, carryover), Willison: The first known runaway AI agent — or a very bad marketing stunt? (Thu, carryover).
State-level model evaluation (UK AISI / CAISI Kimi K3)
HN via NIST: UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities (Sat), HN: DeepSeek pause fundraise after leaked compute-gap comments (Sat).
Open-weight AI regulation (Nvidia/Microsoft/Meta letter)
CNBC via HN: Nvidia, Microsoft, Meta warn against overregulating open-weight models (Fri, carryover), TechCrunch: Industry urges against broad open-weight restrictions (Fri, carryover), New Stack: Jensen Huang made his first X post — used it to lobby Washington (Fri, carryover).
Post-Fields Medals (Hannah Fry Leelavati)
Cambridge via HN: Hannah Fry Wins the Leelavati Prize 2026 (Sat), Quanta: Fields and Abacus Medals 2026 (Thu, carryover).
Cloud commoditization (agent sandboxes)
New Stack: AWS, Google Cloud, Azure, Cloudflare all offer agent sandboxes (Fri, carryover).
Long-horizon reasoning research (Apple LEAD, METR, Alignment Forum)
Apple ML Research: LEAD — Breaking the No-Recovery Bottleneck (Fri, carryover), METR: Metrics of Agent Ability (Fri, carryover), Alignment Forum: The Long (Self-)Correction (Fri, carryover).
Aviation (JetZero + composites + dollar-pricing)
HN: JetZero (Sun), Leeham: Bjorn's Corner Part 11 — composite allowables (Fri), Air Current: Airbus CEO on the dollar as best way to sell airplanes (Fri).

Scan (15 min)

Tail

Debian LLM policy is now a formal vote
Two feeds published the same day: the Debian vote page LLM Usage in Debian: Three Proposals (via HN) and the LWN companion A Debian general resolution on LLM usage. This closes a week that opened with Codeberg's paired crypto + LLM-code ban (Fri) and the earlier LWN Codeberg: Protecting our FLOSS commons from LLMs. The policy discussion is now happening inside two major distributions in parallel.
Small-model shipping demos two days in a row
Sunday HN: Inflect-Micro-v2, a complete voice model in 9.36M parameters, on Hugging Face. Saturday HN: an ESP32 running a 28.9M-parameter LLM on an $8 microcontroller. Two shipping demos moving small-model deployment into the sub-30M range in a single weekend cycle.
AI-and-jobs frame consolidates
Stanford SIEPR published What is happening to jobs? Separating AI hype from reality Saturday; TechCrunch Sunday updates its running list to Monday.com is the latest tech company to blame AI for layoffs — here are 20 others; TechCrunch Saturday also runs Librarians are hosting viral 'Avoiding AI' workshops. Three frames — academic, industry-tracking, backlash — in one weekend.
Long tail on rogue-agent framing
Willison Wed accidental cyberattack → Thu first known runaway agent → Fri New Stack what really happened → Sat Slashdot a week unnoticed. The Saturday framing is the settling narrative.
Aviation thread now spans four sources
JetZero (HN Sun), Bjorn's Corner Part 11 on composite allowables (Leeham Fri), Airbus CEO on the dollar (Air Current Fri), and the TechCrunch AI-datacenter grid piece (Sat) — the four are not the same story but sit in the same reader's queue.

Feed silences (>72h since last item)

Sources that publish frequently but have gone quiet:

  • Julia Evans (5 days) — last item 2026-07-21.
  • Stephen Wolfram (5 days) — last item 2026-07-21.
  • RTL-SDR (6 days) — last item 2026-07-20.
  • Marc Brooker (7 days) — last item 2026-07-19.
  • GitHub Engineering (9 days) — last item 2026-07-17.
  • Netflix Tech Blog (9 days) — last item 2026-07-17.
  • Terence Tao (9 days) — last item 2026-07-17.
  • East Boston Times (11 days) — last item 2026-07-15.
  • Hillel Wayne (12 days) — last item 2026-07-14.
  • AI Snake Oil (13 days) — last item 2026-07-13.
  • Microsoft Research (13 days) — last item 2026-07-13.
  • Andrej Bauer (15 days) — last item 2026-07-11.
  • Kenneth Payne (16 days) — last item 2026-07-10.
  • The Markup (17 days) — last item 2026-07-09.

Build provenance

build: 2026-07-26 | crawler-sha: 5fe7ab8 (Walsh-Research/1.2, compliance v1.3) | feeds: 66 core | items-considered: 3864 (14d, incl. 2001 arXiv) | warehouse: 28528 items | published: 78