"1,000 days left" Anthropic founder

Credibility score: 71/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.

BSmeter analyzed ""1,000 days left" Anthropic founder" and rated it 71/100 for credibility (a BS score of 29/100 — mostly credible), on 2026-05-11. Its weakest claim — "Humanity approaching AI endgame like frog in boiling water" — scored 50/100 and was flagged as opinion. 30 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.

Claims analyzed

Humanity approaching AI endgame like frog in boiling water — Opinion (50/100)

Classic frog analogy — dramatic but just hype for now 🚩

Jack calls autonomous AI R&D a historic big deal — Solid (75/100)

👌

RSI and intelligence explosion coming online soon — Opinion (50/100)

Hype train leaving the station — classic AI doomer vibes 🚂😬

METR chart shows agent capabilities exploding like Wait But Why ASI graph with Claude Mythos — Solid (80/100)

👌📈

Google DeepMind hired Alex Imas as Director of AGI Economics this week — Verified (95/100)

✅ Fresh hire, nailed it

Shane Legg: economy is trading labor for resources since hunter-gatherer days — Just Vibes (50/100)

Shane dropping econ 101 with AGI twist — solid framing 💡

Google DeepMind hiring for AGI economy disruption prep — Solid (85/100)

👌

Jack Clark pondering AGI's impact on jobs and economy — Solid (80/100)

✅

AGI transition is humanity's wildest ever — Opinion (50/100)

Classic AI hype — wildest? Industrial Rev says hold my beer 📈😏

Automated AI R&D crosses point of no return — Opinion (50/100)

Dramatic flair but fair warning on the unknown 👌

Essay author plans most of 2026 on AI implications — Solid (75/100)

👌

Sci-fi writers say hard to write smarter characters — OK (65/100)

Names mangled but concept tracks ⚠️

Writing superintelligent characters is hard due to lacking cognitive access — Opinion (50/100)

Fair point on author limits — Scott Adams nod checks out 👌

Fully automated AI researcher backed by tons of papers and demos — Solid (80/100)

✅

AI research mostly hypothesis, code, run, results — pretty simple — OK (65/100)

Simplified but directionally right — skips the hard parts ⚠️

AlphaEvolve notable for AI self-improvement — Verified (90/100)

AlphaEvolve spot on — DeepMind's algo evolver 👌✅

Terence Tao is the world's top pure mathematician — Opinion (50/100)

Fair take — guy's a legend, but 'number one' is subjective 👌

Tao said AI hit inflection for math discovery end of 2025 — Solid (85/100)

✅

AlphaEvolve optimizes next-gen TPUs and AI infrastructure — Verified (95/100)

👌💯

AlphaEvolve improved Gemini training and cache policies — Solid (90/100)

✅

Claude Mythos Preview scores 93.9% on SWE-bench — Verified (95/100)

✅

Banks warning about Claude Mythos as cyber risk — Solid (85/100)

👌

Mythos cybersecurity skills are emergent property — Solid (80/100)

✅

Mythos is one of largest models, bigger than Opus — OK (65/100)

Plausible but 'one of the largest' is vague flex ⚠️

AI good at core science skills: direction, experiments, sanity checks — Solid (80/100)

👌

AI improving at long-horizon tasks autonomously — Solid (75/100)

👌

AI mistakes tolerable in research/coding; net positive despite errors — Personal Story (65/100)

Fair take from experience — errors suck but output wins ⚖️

Alpha Evolve uses Google's Gemini LLM with a harness as an agent — Solid (85/100)

👌

Alpha Evolve has prompt samplers, evaluator pools, program database, centered on LMs — Solid (80/100)

✅

AI replicating ML papers via Core Bench is key to science — OK (60/100)

Replication is science 101, AI trying it via Core Bench? Promising but unproven at scale ⚠️

See the full analysis with sources and timestamps →