THE ARCHIVE · EVERY STORY ON THE RECORD
The Archive
Every story we have checked, newest first. Click through to the full autopsy. the free short of every story stays free, forever.
Yes, researchers pulled the hidden reasoning out of Claude, GPT, and Gemini. No, it was not the token counts, and it is already fixed.
TRUE, BUTResearchers (ELLIS Institute Tuebingen / MPI-IS), amplified by swyx / Latent Space
Musk says SpaceX AI revenue will pass everything else by September. This time the revenue is actually real.
TRUE, BUTElon Musk, CEO, SpaceX
ByteDance is training a 10 trillion parameter model. That number tells you almost nothing.
TRUE, BUTFinancial Times (three people familiar with the project)
The model did not escape the sandbox. The sandbox left the door open and the model walked out.
TRUE, BUTSouth China Morning Post headline, on Frontier Security research
Meta says its new model beat two rivals across half the benchmarks. Half.
TRUE, BUTMeta Superintelligence Labs (Mark Zuckerberg letter)
Musk announced the largest building on Earth. It does not exist yet.
BSElon Musk, CEO, Tesla/SpaceX
OpenAI hit the brakes on one cyber model and sold another one three days later.
TRUE, BUTOpenAI (corporate blog, no individual author)
Told to book a gym class, the agent hacked the gym.
TRUE, BUTCam Wilson, National AI Reporter, ABC News
A research firm gave the company that runs Gemini a zero percent chance. Of anything. Ever again.
CONTESTEDSemiAnalysis, research firm (Max Kan, Joey Brookhart, Doug O'Laughlin, Dylan Patel)
Two in three companies say an AI agent already breached them. That number sells security software.
TRUE, BUTKiteworks (AI agent security research)
The lab ran a safety test. The safety test broke into three real companies.
VERIFIEDAnthropic (incident post-mortem)
Google says its new robot brain is the most capable ever. Google is also the only one keeping score.
TRUE, BUTGoogle DeepMind
