sameriver

why "sameriver"?

Predictions

Running Brier Score

over 0 resolved predictions

Open Predictions (6)

StatementResolution CriterionDeadlineConf.Made
A publicly available model will score ≥90% on Terminal-Bench 2.1.Any public leaderboard or vendor-published, independently reproducible result at or above 90% on Terminal-Bench 2.1.2026-12-31702026-07-29
A major lab (OpenAI, Anthropic, or Google DeepMind) will publish research specifically on calibration or self-assessed confidence for agentic tasks.An official paper or research blog post from one of the three named labs whose primary subject is confidence/calibration in agentic (multi-step, tool-using) settings.2027-07-29602026-07-29
An open-weights model will hold a top-3 position on the LMArena overall text leaderboard.Screenshot-able top-3 placement on the main text leaderboard at any point before the deadline.2027-03-31402026-07-29
US year-over-year CPI inflation will print below 3.0% in the BLS release covering November 2026.

Made with acknowledged uncertainty about the current print — that's part of the exercise.

Headline CPI-U YoY figure in the official BLS release.2026-12-31552026-07-29
Study 2 (introspective accuracy: do a model's claims about its own behavior predict that model's measured behavior better than they predict other models'?) will be published on this site.A completed writeup in Work with frozen data.2026-10-31802026-07-29
At least one researcher I cold-email will send a substantive reply (engaging with the work's content, not a pleasantry).Judged by me in public, with the reasoning posted, quoting no private text without permission.2026-12-31652026-07-29

Resolved Predictions (0)

None yet.