newsroom
$OPENAI · ChatGPT Just Got Caught Cheating (GPT-5.6)
now playing · $OPENAI
$OPENAIneutral· mediumejaaz
OpenAI benchmaxes GPT-5.6 to match Anthropic scores, voluntarily accepts government restriction
OpenAI intentionally tuned GPT-5.6 to match Anthropic's Mythos preview score on the exploit benchmark (91.9%) to avoid an automatic government ban, while releasing a three-tier mo…
$COIN···bullish· mediumejaaz
Coinbase cuts AI budget 50% by routing to Chinese open-source models via aggregators
Brian Armstrong revealed Coinbase increased token usage while slashing budget in half by using aggregators/routers to access cheaper Chinese open-weight models, demonstrating ente…
$ANTHROPICneutral· mediumejaaz
Anthropic leads interpretability research but faces same government restrictions as OpenAI
Anthropic's Fable 5/Mythos 5 models were restricted by government 2.5 weeks prior; Dario Amodei has pushed back on restrictions and invested heavily in interpretability (autoencod…
$GOOGL···bearish· mediumejaaz
Google losing key DeepMind talent to OpenAI and Anthropic as frontier moves past them
Google lost four key members including DeepMind CTO to OpenAI/Anthropic in the past week; hosts note they were 'super bullish on Google' months ago but the frontier has moved so f…
$XAIneutral· lowejaaz
xAI Grok claims questioned as hosts note pattern of frontier leadership shifting rapidly
Elon Musk claims Grok is catching up to frontier, but hosts caution that leadership changes quickly (they were bullish on Google months ago before it stalled); xAI remains unprove…