Morning Brief: Monday, June 22

Sixty-four feeds. Two weeks. 5,275 items reduced to what follows. (what we track, how we crawl, subscribe)

Monday. The headline: Sakana ships Fugu; Vercel adds Fugu Ultra to its AI Gateway the same day. A Codex logging bug may write terabytes to local SSDs. A developer forks systemd over a new birth-date field.

Fable arc day 15: an open-weight push continues across feeds — Apertus debuts as an "open foundation model for sovereign AI," GLM 5.2 vs. Opus comparisons surface, and a HN essay argues there is minimal downside to switching to open models. Project Fetch Phase Two still echoes.

Agent security pressure intensifies: The New Stack reports "agentjacking" — a public Sentry key is enough to hijack Claude Code, Cursor, and Codex. China's robotaxi dominance is now formally scoreboarded. Police raid a Danish privacy activist.

Top (5-7 min)

Sakana releases Fugu
HN, 2026-06-22. Sakana publishes Fugu. The Vercel AI Gateway adds Fugu Ultra the same day — a coordinated launch between the model lab and a major routing layer.
Codex logging bug may write TBs to local SSDs
HN, 2026-06-22. Filed against the Codex CLI. The blast radius is consumer hardware; the failure mode is a runaway log. Worth noting how quickly an agent's debug output can become an availability incident.
Someone forked systemd over its new birth date field
Slashdot, 2026-06-22. A fork over a single schema addition. The substantive question — what timestamps belong in init — sits underneath a familiar governance dispute.
Agentjacking: a public Sentry key hijacks Claude Code, Cursor, Codex
The New Stack, 2026-06-21. A leaked Sentry DSN, combined with an MCP server, is enough to take over agentic developer tools. The exposure surface follows from giving an LLM outbound HTTP plus shared credentials.
New robotaxi scorecard shows China's dominance
TechCrunch, 2026-06-21. TechCrunch Mobility publishes a comparative scorecard. The methodological framing matters as much as the result.
Project Fetch: Phase Two
HN, 2026-06-21. Anthropic's research output continues through the Fable/Mythos political fallout. Carryover from Sunday but still active across feeds.
Danish privacy activist Lars Andersen raided by police
HN, 2026-06-22. Posted to Twitter; HN front page. State action against a privacy advocate inside the EU.

Themes this week

Fable/Mythos arc (day 15)
The arc: launch (06-09), invisible guardrails (06-10), Anthropic apologizes (06-11), proactivity bias (06-12), government suspension (06-13), geopolitical fallout (06-14), D.C. cleanup (06-15), cyber defense protest (06-16), red-team study on arXiv (06-17), employees accuse admin / sales up (06-18), prylint, Open Fable forecast (06-19), Schneier on Fable, PGP to Mythos: export controls don't work (06-19/20), open-weight case strengthens, Whittaker: chatbots not your friends, Project Fetch Phase Two (06-20/21), day 15: minimal downside to switching to open models, Apertus: open foundation model for sovereign AI, GLM 5.2 vs. Opus, who benefits from a crackdown on Anthropic (06-21/22).
Agent security
agentjacking via Sentry MCP keys, Codex log bug writes TBs to local SSDs, identity verification on Claude, Deno Desktop, Rust ecosystem AI security engineer in residence, patterns for cybersecurity evals. The agent attack surface is now: third-party credentials, runaway logging, identity bootstrap.
AI governance and surveillance
US bill: AI chip location tracking, UK on VPNs and teen chatbot restrictions, Guardrails Alliance super-PAC, cops arrested for stalking via Flock cameras, facial recognition on Kansas City buses, Danish privacy activist raided. Surveillance continues to expand both in scope (municipal cameras, biometric transit) and in enforcement against critics.
Open-weight push
Sakana Fugu, Fugu Ultra on Vercel AI Gateway, Apertus, GLM 5.2 vs. Opus, minimal downside switching to open models, local fine-tuning Qwen 3:0.6B. Open weights, sovereign foundation models, and local fine-tuning are showing up across feeds in the same week — the Fable arc is shaping the market, not just the discourse.

Scan (15 min)

Tail

Feed silences (diagnostic)

  • arxiv-cs-ai: 0 items on 06-20, 06-21, 06-22 (weekend + Monday morning, expected); 289 on 06-19; 3,391 in 14-day window.
  • anthropic-generated: last item 06-17 (Seoul office).
  • claude-code-releases: v2.1.185 (06-20), latest release.
  • deepmind-blog: silent since 06-01.
  • Apple ML Research: last item 06-08.
  • Ink & Switch: posted 06-19 ("Artificial"), then silent.
  • Microsoft Research: last item 06-12.
  • Google Research: last item 06-16.

Build provenance

build: 2026-06-22 | crawler-sha: c04c1d1 (Walsh-Research/1.2, compliance v1.3) | feeds: 64 core | items-considered: 5275 (14d, incl. 3391 arXiv) | warehouse: 17183 items | published: 49 | note: Fable arc day 15 (Apertus, GLM 5.2 vs Opus, "minimal downside" essay, Project Fetch Phase Two echo); Sakana Fugu + Fugu Ultra coordinated launch; Codex SSD log bug; systemd birth-date fork; agentjacking via Sentry MCP; China robotaxi scorecard; Danish privacy activist raided