▲AI Hype Cycles and Funding Narratives

Score history

Daily conviction score, 0 to 100. Higher means the thesis is more strongly corroborated.

Now 100

Showing the last few days. Unlock full score history.

Why the conviction moved

• Aug 28
  Strengthened +5

  Sam Altman said OpenAI will have an internal system he calls AGI before the end of 2026, resting on OpenAI's own definition and pointing to Astra, a model that has not been released or independently evaluated. An unfalsifiable capability claim defined by the claimant and attached to an unreleased system is precisely the rhetoric-versus-verifiable-capability gap the thesis tracks, and it arrives while OpenAI is raising against enormous capex commitments.

• Aug 27
  Strengthened +2

  ESQ-Bench shows headline text-to-SQL accuracy above 89 percent collapses against 465-table production Oracle schemas, with failures that run cleanly and return wrong results. It gives buyers a concrete instrument for separating benchmark claims from deployable capability in one of the most commercially marketed agent categories.

• Aug 27
  Strengthened +3

  OpenAI's Jalapeño-versus-GB300 numbers come from a benchmark suite the company selected and executed itself, with no third-party replication. Vendor-run comparisons on self-chosen suites are exactly the class of claim the market has been discounting, so the disclosure functions as narrative capital until independent runs exist.

Showing the last 2 days. Unlock the full record.

Source trail

• Supporting · August 28, 2026

  Altman Says OpenAI Will Have an Internal System He Calls AGI Before the End of 2026

  Sam Altman said OpenAI will have an internal system he calls AGI before the end of 2026, resting on OpenAI's own definition and pointing to Astra, a model that has not been released or independently evaluated. An unfalsifiable capability claim defined by the claimant and attached to an unreleased system is precisely the rhetoric-versus-verifiable-capability gap the thesis tracks, and it arrives while OpenAI is raising against enormous capex commitments.

  AI Post (Telegram)

• Supporting · August 27, 2026

  A New Benchmark Argues Text-to-SQL Accuracy Above 89 Percent Does Not Survive Real Enterprise Schemas

  ESQ-Bench shows headline text-to-SQL accuracy above 89 percent collapses against 465-table production Oracle schemas, with failures that run cleanly and return wrong results. It gives buyers a concrete instrument for separating benchmark claims from deployable capability in one of the most commercially marketed agent categories.

  arXiv cs.AI