Morning Brief: Saturday, July 11

Sixty-five feeds. Two weeks. 4,181 items reduced to what follows. (what we track, how we crawl, subscribe)

Saturday. Latent Space files the honest headline of the day: not much happened today. But two stories do land. First, Apple sues OpenAI, accusing ex-employees of stealing trade secrets — a legal counter to Friday's launch blowout, with Slashdot picking up the Apple/OpenAI angle and TechCrunch running the trade-secret theft suit. Second, China lands a rocket during an orbital launch for the first time. On the security beat: US cybersecurity agency CISA reveals it had to build its incident playbook during the incident. Pluralistic argues workplace flexibility isn't. Yesterday's carryover still warrants attention: HN surfaced a paper claiming GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture; The New Stack triangulated the OpenAI/Microsoft/Anthropic split on who runs the agent; Apple ML Research formalized behavioral privacy leakage in agentic negotiation (well-timed given today's suit). Weekend arXiv batch: Harness Engineering for Auditable Enterprise LLM Agents (directly dogfoodable), Overthinking: amplifying reasoning weights to extract learned secrets, and Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents.

Top (5-7 min)

Apple Sues OpenAI, Accusing It of Stealing Company Secrets: Slashdot, 2026-07-11. The morning-after answer to Friday's launch. Slashdot's frame is the corporate suit; TechCrunch and 9to5mac carry the trade-secret ex-employees angle. Read against yesterday's Apple ML Research paper on behavioral privacy leakage in agentic negotiation — the timing is not accidental.
Apple sues OpenAI over alleged trade secret theft: TechCrunch, 2026-07-10. The TC filing note: the suit names ex-Apple employees who moved to OpenAI as the vector. First-mover legal escalation between the two companies most likely to matter to the Work superapp endgame.
China Lands Rocket During an Orbital Launch For First Time: Slashdot, 2026-07-11. First orbital-class landing outside of SpaceX. Adjacent to yesterday's US-China AI decoupling thread (Nobel chemist to China, lawmakers probe Chinese AI models) — the industrial-capacity story keeps compounding.
AINews: not much happened today: Latent Space, 2026-07-11. The honest Saturday headline after Friday's OpenAI blowout. Read as a signal: the launch cycle has emptied the news pipeline.
US cybersecurity agency CISA had to build its incident playbook during the incident: TechCrunch, 2026-07-11. The security beat's confession: the national incident-response playbook did not exist when it was needed. Reference for anyone building agentic incident-response.
GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture: HN, 2026-07-10. Carryover worth flagging: OpenAI-hosted PDF claiming a proof of a 60-year-old open problem in graph theory. HN is doing the review; independent verification still pending as of this brief.
Pluralistic: Workplace "flexibility" isn't: Pluralistic, 2026-07-11. Doctorow's Saturday piece: employer-defined flexibility as risk transfer. Companion to yesterday's rights for robots post.

Themes this week

Apple v. OpenAI opens a legal front: Slashdot: Apple sues OpenAI, stealing company secrets (today), TC: Apple sues OpenAI over alleged trade secret theft (today), HN: Apple sues OpenAI, ex-employees stealing trade secrets (Fri), Apple ML: behavioral privacy leakage in agentic negotiation (Fri).
OpenAI's Friday launch — the carryover: Latent Space: GPT-5.6 Sol/Terra/Luna, Codex becomes ChatGPT superapp (Fri), HN: GPT-5.6 Sol Ultra produces proof of Cycle Double Cover Conjecture (Fri), TC: NYT says OpenAI hid evidence in ChatGPT copyright trial (Fri), OpenAI: Deutsche Telekom rewiring telecommunications with AI (Fri).
Who runs the agent — control planes: TNS: OpenAI, Microsoft & Anthropic agree on who runs the agent, disagree on what you can take back (Fri), TNS: enterprise AI benchmarks are broken (Fri), TNS: why retrieval quality is becoming the defining challenge (Fri).
Alignment theory and interpretability: Alignment Forum: value generalisation — value correction (Fri), Alignment Forum: how robust are NL autoencoders to initialization (Fri), MIT TR: Anthropic found a hidden space where Claude puzzles over concepts (Fri).
Security beat — playbooks, doorbells, and scrapers: TC: CISA had to build its incident playbook during the incident (today), Slashdot: Russia hacks doorbell cameras to spy on NATO bases (Fri), LWN: update on the scraper situation (Fri), Schneier: AI surveillance and social progress (Fri).
China industrial capacity and AI decoupling: Slashdot: China lands rocket during orbital launch for first time (today), Slashdot: Nobel-winning US chemist to China to lead AI institute (Fri).

Scan (15 min)

Saturday signal (short list — quiet day)
- Apple Sues OpenAI, Accusing It of Stealing Company Secrets, Slashdot, 07-11
- China Lands Rocket During an Orbital Launch For First Time, Slashdot, 07-11
- AINews: not much happened today, Latent Space, 07-11
- US cybersecurity agency CISA had to build its incident playbook during the incident, TechCrunch, 07-11
- Phia accused of 'cookie stuffing', TechCrunch, 07-11
- Pluralistic: Workplace "flexibility" isn't, Pluralistic, 07-11
arXiv weekend batch (cs.AI, 216 items today)
- From Prompts to Contracts: Harness Engineering for Auditable Enterprise LLM Agents, arXiv cs.AI, 07-11
- Overthinking: Amplifying Reasoning Weights to Extract Learned Secrets, arXiv cs.AI, 07-11
- Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents, arXiv cs.AI, 07-11
- Context Graphs for Proactive Enterprise Agents, arXiv cs.AI, 07-11
- Game Theory Driven Multi-Agent Framework Mitigates Language Model Hallucination, arXiv cs.AI, 07-11
- CausalDS: Benchmarking Causal Reasoning in Data-Science Agents, arXiv cs.AI, 07-11
Hackaday build log
- A Look Inside a 1997 BBC Ceefax Generator, Hackaday, 07-11
- Fixing the Fix for a 3dfx Voodoo Card's Overly Bright Picture, Hackaday, 07-11
- Reject Modernity, Return to 80s, Learn C, Hackaday, 07-11
Yesterday's carryover worth re-reading Saturday morning
- GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture, HN, 07-10
- OpenAI launches GPT 5.6 Sol/Terra/Luna, Latent Space, 07-10
- Behavioral Privacy Leakage in Agentic Negotiation, Apple ML Research, 07-10
- OpenAI, Microsoft & Anthropic disagree on what you can take back, The New Stack, 07-10
- Enterprise AI benchmarks are broken, The New Stack, 07-10
- The Ambulatory Intelligence Gap, Databricks, 07-10
- Value generalisation: value correction, Alignment Forum, 07-10
- Anthropic found a hidden space where Claude puzzles over concepts, MIT Tech Review, 07-09
- Scarf moves away from Haskell after 7 years, HN, 07-10
- AI-generated videos to maximally drive a target brain region, HN, 07-10

Tail

Claude Code v2.1.207, claude-code-releases, 07-11
The vintage beauty of Soviet control rooms (2018), HN, 07-11
Almost $1B Later, the US Still Can't Make a Medical Glove, HN, 07-11

Feed silences (diagnostic)

arxiv-cs-ai: 216 items today — the weekend catch-up landed as expected after Friday's 0-item cliff. Monday will see the next arXiv reset.
Anthropic first-party, OpenAI, Simon Willison, Apple ML Research, Google Research, ClickHouse, EFF Deeplinks, The New Stack, LWN, Planet Clojure, Schneier, GitHub Blog, Interconnects, Kenneth Payne, Martin Fowler, Microsoft Research, Cloudflare, Charity Majors, Databricks, Hugging Face Blog, MIT Technology Review, 404 Media, Nature Machine Intelligence, Alignment Forum, Quanta Magazine, Grafana Labs: no primary posts today — the expected Saturday silence across most first-party channels.
Netflix Tech Blog, James Bornholt: persistent DNS/TLS errors continue (unchanged).
WBUR: no Endless Thread today.
bitsavers (6 feeds): connected, 0 items (sparse archive, expected).
claude-code-releases: v2.1.207 shipped today.

Build provenance

build: 2026-07-11 | 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: 4181 (14d, incl. 2397 arXiv) | warehouse: 24254 items | published: 7 | note: Saturday. Quiet day per Latent Space's own headline. Fresh signals: Apple sues OpenAI over trade-secret theft (ex-employees named), China lands orbital rocket for first time, CISA reveals it built incident playbook during incident. Carryover: GPT-5.6 Sol Ultra Cycle Double Cover proof (unverified), OpenAI/MS/Anthropic control-plane split (TNS), Apple ML behavioral privacy paper — well-timed against today's suit. arXiv weekend batch: harness engineering for auditable enterprise LLM agents, Overthinking attack extracts learned secrets by amplifying reasoning weights.