trader score
·
hit rate
·
mean α
·
signals 14d
28
Grade = how their written analysis reads (A best). Trader score = how their last-20 timestamped calls performed vs SPY.
Their picks, scored
Not yet in the scored-bets universe (fewer than 20 scoreable calls).
Recent signals15receipts included
| date (PT) | ticker | author | sent | what they said | since then | receipt |
|---|---|---|---|---|---|---|
| 2026-10-15 | · | @OfirPress | · | Notes growing use of ProgramBench to study multi-agent swarms. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Clarifies that two benchmark/data setups are nearly identical apart from volume. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Shares the SWE-bench Multimodal leaderboard. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Endorses another benchmark as likely to advance the field, though details require linked context. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Explains that multimodal coding benchmarks test coding, image processing and sometimes 3D vision. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Announces the open-source launch of a benchmark project and identifies its contributors. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Reports launch of a SWE-bench Multimodal leaderboard for visual software bugs. | · | tweet ↗ |
| 2026-10-14 | · | @OfirPress | · | Describes a method enabling language models to explain chess moves through generated board-analysis code. | · | tweet ↗ |
| 2026-10-13 | · | @OfirPress | · | Argues frontier AI needs benchmarks formalizing the gap between agent capability and human needs. | · | tweet ↗ |
| 2026-10-13 | · | @OfirPress | · | Proposes benchmark tasks that are long-horizon, real-world and scalable to collect from the web. | · | tweet ↗ |
| 2026-10-13 | · | @OfirPress | · | Says many human tasks are now too easy for AI benchmarks and introduces new benchmark-building guidelines. | · | tweet ↗ |
| 2026-10-12 | · | @OfirPress | · | Author mocks AGI claims based on the need for layered agent supervision. | · | tweet ↗ |
| 2026-10-12 | · | @OfirPress | · | Author discusses benchmark robustness, LM-as-judge limitations, and possible post-training use. | · | tweet ↗ |
| 2026-10-12 | · | @OfirPress | · | Praises strong ProgramBench results. | · | tweet ↗ |
| 2026-10-12 | · | @OfirPress | · | Endorses research using ProgramBench and asks whether generated mutants resemble real bugs. | · | tweet ↗ |
Grade is our human read-worthiness rating; trader score is a rolling 20-bet hit-rate/alpha composite — different things, often disagreeing. “Since then” is direction-unaware in the table; the summary line above adjusts for which way they leaned.