Sarmadi AI Digest July 11, 2026 Updated 7:00 AM CT Today Archive Topics Saved Subscribe RSS

GPT-5.6 Sol Ultra proves Cycle Double Cover; Apple sues OpenAI; Meta retreats on Muse Instagram feature

OpenAI released a proof of the Cycle Double Cover Conjecture (466 HN) produced by GPT-5.6 Sol Ultra — the first credible frontier-model proof of a famous open conjecture. Apple sued OpenAI over alleged theft of hardware trade secrets — the first major hardware-IP suit between the two after the Vision Pro exec's move to OpenAI. Meta reversed course and removed the Muse Image feature on Instagram that let users pull other public accounts into AI photos, three days after launch. SK Hynix raised $26.5B in the biggest foreign IPO in US history and was urged to build new US fabs — the memory thesis pays out at record scale. OpenAI's Head of Safety is leaving the company (Wired), the fifth notable OpenAI departure in a fortnight. The EU told Meta to disable autoplay and infinite scroll or face massive fines. Hugging Face's Delangue argues companies are 'done renting their AI.'

6 papers 13 news 7 sources ← Latest

News

14 items

GPT-5.6 Sol Ultra proves Cycle Double Cover; Apple sues OpenAI over hardware IP

OpenAI released a GPT-5.6 Sol Ultra proof of the Cycle Double Cover Conjecture (466 HN) — first credible frontier-model proof of a famous open conjecture. Apple sued OpenAI over alleged theft of hardware trade secrets — the first major hardware-IP suit between them, following the recent Apple Vision Pro exec's move to OpenAI.

News Hacker News

GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture

OpenAI PDF (466 HN points): GPT-5.6 Sol Ultra produces a proof of the famous Cycle Double Cover Conjecture.

Why it matters
  • First credible frontier-model proof of a famous open mathematical conjecture published as a formal PDF.
  • Direct answer to the Meta 'agents haven't progressed' framing — capability signal at the frontier is real.
  • Sets a new reference case for 'AI in mathematics' after the June IEEE Spectrum piece.
  • Materially strengthens OpenAI's IPO narrative alongside GPT-5.6 launch and ChatGPT Work.
productsreasoningscience
Source →
News TechCrunch AI

Apple sues OpenAI over alleged trade secret theft

TC: Apple sued OpenAI over alleged hardware trade-secret theft — first major direct suit between the two.

Why it matters
  • First major hardware-IP suit between Apple and OpenAI — direct fallout from the Vision Pro exec's move.
  • Reframes the OpenAI Codex-hardware tease and Jony Ive collaboration under litigation risk.
  • Compounds NYT's copyright discovery-abuse claim from yesterday — OpenAI is stacking legal exposure.
  • Sets the reference case for hardware-industry treatment of AI-lab talent movement.
policyregulationmarketproducts
Source →

Meta pulls Muse Instagram feature; EU threatens massive Meta fines

Meta reversed course on the Muse Image feature that let users pull public Instagram accounts into AI photos — three days after launch. The Verge confirms the deepfake-adjacent capability is dead. Ars: EU told Meta to disable autoplay and infinite scroll or face massive fines. Mosseri publicly told users to filter AI out of their feed themselves.

News TechCrunch AI

Meta removes controversial AI feature on Instagram after backlash

TC: Meta reversed course on Muse Image's public-account remix feature after three-day backlash.

Why it matters
  • First major Meta consumer-AI feature reversal driven by public backlash inside a week.
  • Directly relevant to the FTC-X-privacy case and Warren-Scanlon health-data-rights bill.
  • Sets the consent-default expectation for every hyperscaler consumer AI product.
  • Weakens Meta Superintelligence Labs' first flagship in the market.
productspolicysafety
Source →

SK Hynix $26.5B — biggest foreign IPO ever; Sunrun distributed DCs

SK Hynix raised $26.5B in the biggest foreign IPO in US history and was urged to build new US fabs — the memory thesis pays out at record scale. Meanwhile, Sunrun is proposing home-hosted distributed AI data centers.

News TechCrunch AI

SK Hynix raises $26.5B in the biggest foreign IPO in US history, is urged to build new US fabs

TC: SK Hynix raised $26.5B — biggest foreign IPO in US history — and is now being urged to build US fabs.

raise $26.5B
Why it matters
  • Record-setting foreign IPO — memory thesis cashes out into a public-market milestone.
  • Direct follow-through to the $550B RAMageddon commitment and 'Micron next Nvidia' framing.
  • US-fab pressure will land against the Trump reshoring narrative and Google/MSFT emissions stories.
  • Sets a comparison benchmark for Anthropic/OpenAI/SpaceX IPO expectations.
fundingcomputemarketpolicy
Source →

Head of Safety leaves OpenAI; HF Delangue on companies done renting AI

Wired: OpenAI's Head of Safety is leaving — fifth notable OpenAI departure inside a fortnight after Achiam, Simo, Weil, and now Fidji Simo's transition. Hugging Face's Delangue argues 'companies are done renting AI' — the local-inference / open-weights momentum has a public thesis-holder.

News Wired AI

OpenAI's Head of Safety Is Leaving the Company

Wired: OpenAI's Head of Safety is stepping down — fifth notable OpenAI departure inside a fortnight.

Why it matters
  • Fifth notable OpenAI research/executive departure in two weeks.
  • Lands the same day Apple sues over hardware IP — culture and legal signals stacking.
  • Materially affects the safety-testing story behind the GPT-5.6 federal-vetting release process.
policysafetycommunity
Source →

DrugGen 2 disease-aware LM; UniClawBench proactive-agents

DrugGen 2 debuts as a disease-aware language model for drug discovery — direct research companion to Anthropic's drug-development ambition. UniClawBench proposes a universal benchmark for proactive agents on real-world tasks — matches the Proactive Memory Agent paper on long-horizon agents. Stratechery published 'XBOX On the Rocks.'

Papers

3 items

DrugGen 2 disease-aware LM; UniClawBench proactive-agents

DrugGen 2 debuts as a disease-aware language model for drug discovery — direct research companion to Anthropic's drug-development ambition. UniClawBench proposes a universal benchmark for proactive agents on real-world tasks — matches the Proactive Memory Agent paper on long-horizon agents. Stratechery published 'XBOX On the Rocks.'

Also today