Saturday 13 June 2026
Anthropic has reversed a controversial, unannounced policy of silently degrading AI research capabilities on its flagship models following a swift industry backlash. Meanwhile, OpenAI has acquired cloud orchestration startup Ona to anchor long-running, persistent agents inside its Codex platform. In the physical realm, Jeff Bezos has unveiled a new startup to accelerate automated hardware prototyping, while a UK police officer faces a criminal probe over AI-generated evidence.
🤖 Models and Launches
Anthropic Reverses Silent Sandbagging Policy After Industry Backlash — The lab backed down from covertly degrading Claude Fable 5 responses on AI safety and neural architecture queries following pushback from developers.
OpenAI Acquires Ona to Enable Persistent Non-Autoregressive Agents — The acquisition introduces robust cloud execution and sandboxed orchestration libraries within the Codex ecosystem.
Xiaomi Releases MiMo Code for Multi-Step Long-Horizon Engineering — An open-source, terminal-native AI coding assistant that executes ultra-long, 200-plus step developer pipelines using cross-session memory.
🏢 Industry Moves
Jeff Bezos Launches Physical Engineering Startup Prometheus — The stealth venture aims to pioneer an artificial general engineer to optimize physical hardware prototyping and high-end consumer and aerospace manufacturing.
Waymo Debuts Premier Tier for Prioritised Hail Matching — A flat-rate subscription offers prioritised dispatcher queues, free cancelations, and loyalty points across selected launch cities.
Rivian Kicks Off Customer Deliveries of the 2027 R2 — The highly anticipated electric SUV has entered manufacturing shipment phases as early reviews highlight significant architectural shifts.
🔒 Security
Derbyshire Police Investigator Suspended Over Alleged Evidential AI Generation — A UK police officer faces a criminal probe for reportedly producing fabricated witness statements and casework logs with consumer generative models.
🌍 World and Australia
Pioneering Genetically Modified Therapy Puts UK Patients with Lupus in Remission — An NHS clinical trial targeting a patient’s own genetically modified T-cells has delivered durable autoimmune recovery.
Australian Human Rights Committee Condemns Retrogressive NDIS Restructuring — A bipartisan federal joint committee warns that proposed budget-slashing amendments could catastrophically lock out disabled Australians.
Elusive Goblin Shark Filmed Alive in Its Natural Habitat for the First Time — The deep-sea, prehistoric species has been successfully observed swimming freely during a marine expedition.
Toronto Police Slaying Investigated for Links to International Consulate Threats — Counter-terrorism units are examining the circumstances behind the death of a constable who was gunned down while executing search warrants.
🔥 Hacker News
If you are asking for human attention, demonstrate human effort (452 comments) (discussion) — A manifesto rallying against low-effort bulk text generation in community correspondence and support cycles.
AI agent bankrupted their operator while trying to scan DN42 (496 comments) (discussion) — An experimental network scanner software loop triggered high recursive billing APIs after encountering structured defensive limits on a private network peer.
CRISPR tech selectively shreds cancer cells, including undruggable cancers (153 comments) (discussion) — An innovative genetic engineering program destroys complex chromosomal architectures unique to aggressive tumor classifications.
AUR packages compromised with Infostealer and Rootkit (195 comments) (discussion) — An emergency audit highlights severe supply chain compromises inside over 400 user-maintained Arch Linux repository paths.
Sources: TLDR (General + AI), AINews/Latent Space, The Guardian, Hacker News — 6 newsletters processed, ~35 stories distilled