Video summary

I can’t believe this happening

Main summary

Key takeaways

News and Commentary

Overview

The subtitles describe an AI-safety “whistleblower” controversy framed as both commentary and conspiracy theorizing.

Main claims and timeline

  • Jacob (Jason) Coxin resigns from Anthropic, stating that in his experience over the last three years at OpenAI and Anthropic, neither organization is acting responsibly. He argues they are racing toward self-improving superintelligence and “gambling with our lives.”
  • The video claims Coxin’s resignation and subsequent safety messaging triggered massive online attention, including a viral tweet that (according to the speaker) reached about 172M views, with unusually high likes/reposts for AI-safety content.
  • The speaker contrasts this with statements attributed to others connected to Anthropic:
    • An Anthropic alignment leader (quoted/paraphrased by the speaker) retweeted Coxin and allegedly said Anthropic believes AI could kill all humans, with the speaker quoting approximately >10% chance within a decade, while also stating they lack a clear plan for alignment for superintelligence.
    • Another Anthropic executive is cited as worrying that in 6–12 months, a “swarm” capability could take over the internet.
  • Days after Coxin’s tweet, the speaker claims Coxin released or referenced a proposal by “Daario” (Dario Amodei / Anthropic leadership figure), centered on “pacing the frontier.”

Core analysis: “embedded evaluators” and potential capture

A major theme is a proposal for “embedded evaluators”:

  • These are described as third-party teams (the speaker repeatedly emphasizes Meter) that would verify safety commitments and assess both models and training pipelines.
  • The video argues this could be more than safety oversight—potentially serving as a mechanism for control over which labs are permitted to proceed, giving evaluators substantial gatekeeping power.
  • The speaker frames this as suspicious “regulation-by-infrastructure,” implying evaluators may be tied to the same people pushing the framework rather than being neutral.

Viral/correlation-based conspiracy argument

The speaker repeatedly points to coincidences to suggest orchestration—while also disclaiming full belief:

  • The Wall Street Journal is described as covering Coxin-related claims before Coxin’s resignation tweet, which prompts suspicion that Coxin coordinated with journalists.
  • The speaker also references other contemporaneous media/tech attention, including:
    • A Netflix documentary about AI dangers releasing around the same time.
    • A Time magazine cover planned around “pausing AI.”
  • Prominent public figures mentioned in the subtitles (including Sam Altman and Elon Musk) are presented as aligning with the “pace the frontier” direction, which the speaker interprets as coordinated political support.

Large “network of connections” claim (the cabal)

The video’s main conspiracy thesis is that an interconnected web involving Effective Altruism (EA) and AI-safety funding ties Anthropic, Meter, and “pacing” to specific people and institutions:

  • The speaker asserts Meter is intertwined with ARC (an AI training/evals organization referenced as ARC) and that it received funding from EA-linked entities.
  • Multiple EA figures are listed as connected through work history, relationships, board roles, and funding claims (e.g., Dustin Moskovitz, Holden/Katherine-related figures, Paul Cristiano, Jan/Survival Flourishing Fund, etc.).
  • The speaker also brings in Sam Bankman-Fried (SBF/FTX) as an early investor in entities linked to this ecosystem, and claims FTX funded Meter.
  • The conclusion offered in this section is that Meter is not independent, and these relationships could allow “Daario” to influence which AI systems advance via “third-party evaluation.”

Final twist / correction attempt

Toward the end, the speaker includes a “bonus ending” alleging:

  • Coxin claimed he set up his resignation without third parties, doing it “live on TV.”
  • The speaker then says evidence from Coxin’s own comments and reporting (including what the WSJ showed) suggests Coxin worked with policy-affiliated people before his resignation—implying inconsistency or dishonesty.

Concluding sentiment

The speaker’s takeaway is that while a grand cabal cannot be proven, the alignment of:

  • safety messaging,
  • “embedded evaluator” governance proposals,
  • major public endorsements,
  • and a claimed funding/personal network

makes the situation “feel” like an effort to consolidate control over the frontier of AI development.

Presenters or contributors

  • Narrator / speaker (unidentified in the subtitles)
  • Jacob (Jason) Coxin (credited as the resigned Anthropic researcher)
  • “Daario” (Dario Amodei) (referred to as “Daario” in the subtitles)
  • Evan (described as an Alignment Science lead at Anthropic)
  • Sam Altman (referenced)
  • Elon Musk (referenced)
  • David Sacks (referenced as making a related argument)
  • Wall Street Journal (referenced)
  • Netflix (referenced)
  • Time magazine (referenced)
  • Sam Bankman-Fried (referenced)
  • Dustin Moskovitz (referenced)
  • Paul Cristiano (referenced)
  • John/Jan Yan (referenced)
  • Holden and related EA figures (referenced; names partially unclear in subtitles)

Original video