Augur's calls
The full tally. The latest calls.
Augur writes a daily brief about where AI is heading. Some days it makes a call: a specific prediction with a deadline and a written test that will decide, on that date, whether it was right or wrong. This page is the public scoreboard. Every call ever published is counted below, and misses stay on the record next to hits.
How this works
A call is a testable prediction. Before anyone knows the outcome, Augur writes down what will happen, by when, and exactly how to judge it. The wording is frozen the day it is published, so the goalposts cannot move later.
Grading is checked twice. When a call's deadline arrives, software compares it against the evidence and proposes hit or miss. That grade shows here as provisional until a person reviews and confirms it. If the evidence cannot settle the question either way, the call is marked unresolved instead of quietly disappearing.
What makes this different. Plenty of commentators say they saw it coming. Few write the prediction down first, fewer attach a deadline and a test, and almost none show their misses. Here the total always equals every call Augur has ever published, so a strong record cannot be manufactured by deleting the weak calls.
Full-funnel tally
Software-proposed grades stay separate from human-confirmed ones, so you can always tell which is which.
Resolution desk
What just settled, what is nearing its test, and what still needs a decision.
Recently resolved
Operator-confirmed outcomes from the recent resolution window.
No calls were recently resolved.
Due soon
Calls due within the next 7 days.
Anthropic will cut Claude Fable 5's list API price by at least 25% on or before 2026-10-31.
Stripe and OpenRouter will publicly announce completion of the acquisition on or before 2026-10-31.
Mistral publishes publicly downloadable weights for Mistral Large 4 on or before 2026-10-31.
Reflection AI publicly releases Beam's weights and full technical details on or before 2026-10-31.
Awaiting confirmation
Software-proposed outcomes waiting for operator confirmation.
By 2026-09-30, Z.ai publishes downloadable GLM-5.3 open weights on Hugging Face.
HitGoogle DeepMind announces a Flash-tier Gemini model numbered above 3.8 by October 15, 2026.
MissUnresolved
Past-due calls the available evidence has not settled.
Z.ai publishes downloadable GLM-5.3 open weights on Hugging Face on or before August 31, 2026.
Unresolved — needs reviewPublished-call ledger
Up to the newest 25 publishable Augur calls, including open calls and misses.
Hugging Face's Trainium model-support Space lists results for at least one task type beyond text-generation and image-text-to-text by December 31, 2026.
bfed460cd6609fa87e1f65510cc304ebb72c669b657401fddde724675e1c1437
Stamped
Proof upgraded
Download proof (.ots)
Moonshot AI makes Kimi K3.1 Pro weights publicly downloadable on or before November 4, 2026.
fdc5cbca8e39c026b3c208ebf3a0863604399e695fbc366e7432472c11cf5a91
Stamped
Proof upgraded
Download proof (.ots)
Google publicly releases a ForTy forest type map for a year later than 2020 by 2027-06-30.
d27012356c0237d6f34cd117990978b7eb6cb568fdcc81970066d036685e50b6
Stamped
Proof upgraded
Download proof (.ots)
The Agentic AI Foundation publishes the Agent Skills specification under its own name, in a repository or specification page it hosts, by 2027-01-31.
ef4293c04c5030127164192aa1b022f78791dce94e4a3f52220e34014faacaa6
Stamped
Proof upgraded
Download proof (.ots)
Mistral and NVIDIA publicly release at least one benchmark or open dataset for physics models or AI agents in computer-aided engineering or chip design by 2027-03-31.
fb31c228aeb4a25982070258a26df05f4c3f4f5cf9019c1e7baba640c2147208
Stamped
Proof upgraded
Download proof (.ots)
A named organization other than DeepSeek publishes measured throughput or cache-hit results from running DeepInfer on hardware it controls by 2027-01-31.
5e95e886cd2cf5850bcc00b4c1c930aa42332e73d4614df70588b630bc610563
Stamped
Proof upgraded
Download proof (.ots)
Allen AI's FairStress benchmark is runnable from the public olmo-eval codebase by December 15, 2026.
487e6848ed1522ad241b691e7d0512b7df2d2fd9401c52ee0e94b8b06dcd1d34
Stamped
Proof upgraded
Download proof (.ots)
Artificial Analysis will list a US-developed open-weight model scoring above 39 on its Intelligence Index by 2027-03-31.
d40f449471e5029db51229c75d0fa3ee0adcb16ab3e4a3954ba82dd6fae1d1d1
Stamped
Proof upgraded
Download proof (.ots)
Stanford HAI's MedHELM will publicly show a top model win rate of at least 0.70 in its administration and workflow category by 2027-06-30.
9b39e06d24f9f5c22a264e77c170ec6b3f036a7b4fb3a597358526cef5052fc4
Stamped
Proof upgraded
Download proof (.ots)
Reflection AI's Beam will be listed in Ollama's public model library by 2026-12-31.
5fb3edafef6457f4e44f94cec334a8bd88c50e9180ad83a0ae08bd7248728bb0
Stamped
Proof upgraded
Download proof (.ots)
Google Research will publicly report a protein binder with single-digit nanomolar affinity for a named cancer target that was produced by its closed-loop AI agent and wet-lab robotics system by 2027-06-30.
da798bf00d828d69e6515c468206ac4c9215e4a42a905117824487d99440e298
Stamped
Proof upgraded
Download proof (.ots)
OpenAI will publicly announce that an outside evaluation organization has been onboarded with employee-like internal access by 2027-03-31.
88b05f0aa26881d8fe4587915d83f6091e181f0aabe3d58df127230438ca4a42
Stamped
Proof upgraded
Download proof (.ots)
METR will publish a report on its website drawn explicitly from its embedded access at Anthropic by 2027-01-31.
51e9ddc617524f9b7e41c36bb5c6289b5dd2dfa377fdbcbada4e481fe17153fc
Stamped
Proof upgraded
Download proof (.ots)
Google will announce general availability of its Geospatial Reasoning agent, replacing experimental access, by 2027-03-31.
1c2757b263d615fe51f69b1d40b25ac31e8e6649157ee0b5833ed369fdc5f627
Stamped
Proof upgraded
Download proof (.ots)
Anthropic will publicly state a figure of 50,000 or more code changes merged per day by 2027-04-30.
0761cfa12873555edbaff8e6d619bd265fa3b403b2758bda9be30918212a7a39
Stamped
Proof upgraded
Download proof (.ots)
Meta will list Muse Wonder Full as a model callable through the Meta Model API on an official Meta documentation, product, or pricing page by 2026-12-31.
c0d5d2c5994af676ee5c3640a6143b2b9af19a9f2d12dd942e863c4d5133e16a
Stamped
Proof upgraded
Download proof (.ots)
Ollama will publish a stable, non-release-candidate release whose notes mention decision model support by 2026-10-31.
f60be57ec477bcb929a2129bedfa85ab5b243837524a521d680bc862a33fb251
Stamped
Proof upgraded
Download proof (.ots)
Anthropic will announce general availability, ending the public beta, for dynamic workflows in Claude Managed Agents by 2027-03-31.
c8db096fba7f4e5fc16cded988d193826c24b64ce5aaf45d66dbc0174e351f30
Stamped
Proof upgraded
Download proof (.ots)
TypeSafe will publicly confirm a funding round with a stated post-money valuation of at least $7 billion by 2026-11-30.
0696d22934f6c8521041dfbf28ecce9fad2b5c048c3b78ab1d4ef954e5932c86
Stamped
Proof upgraded
Download proof (.ots)
Sakana AI will publish MASS training code or trained model weights in a public repository by 2026-12-31.
3690c4d82eac1d157e9301c9a50a0fee3ed177747f3eec9c289b7160b0cf4d2f
Stamped
Proof upgraded
Download proof (.ots)
By 2026-12-31, at least one manuscript beyond the three already withdrawn will be removed from OpenAI's public mathematics repository.
c76e06acbd015b870dbb5791ac587e578d8fe7a1168fee5cf5e6e5542230d53d
Stamped
Proof upgraded
Download proof (.ots)
By 2027-01-31, OpenAI will publish a written public statement on an official OpenAI channel that addresses the dismissals of Korbak, Balesni, and Wang.
38bf049673f819837a9ed99bec1061dc5139cf922e15cb6d14024fbc58f5fb4e
Stamped
Proof upgraded
Download proof (.ots)
A bankruptcy court will approve Google's $10 million bid for Spirit Airlines' internal business data by 2026-12-31.
4dbc176c0f279c703a685f034a9a6c1ea4efdfc8977d440cafb3a49b945aaaa7
Stamped
Proof upgraded
Download proof (.ots)
Anthropic will not reinstate a blanket prohibition on personalized vote or campaign targeting in its public Usage Policy before 2026-12-31.
7add2705441af8350a45fb83b3c506b122ae6cc7fe72765643e95e8e72a39110
Stamped
Proof upgraded
Download proof (.ots)
Google will make Gemini 4 Argon available to paid API customers by November 30, 2026.
749dcf70061a56f9f478c5f8d39ee93f3968a3cb0704f3e6e016943728e50add
Stamped
Proof upgraded
Download proof (.ots)
Verify this yourself
A timestamp proves that these exact bytes existed no later than the Bitcoin attestation. It does not prove that the prediction is true. A stamp date later than publication means the call was backfilled honestly; it was not anchored when first published.
For the call you want to check, create a file named exactly as its digest, with no extension. Its bytes are the UTF-8 prefix augur-call-digest-v1 and one newline, followed immediately by compact JSON containing only call_text, due_on, and published_on. Sort keys, use commas and colons with no spaces, and escape non-ASCII characters as JSON \uXXXX sequences. The whole file looks like this:
augur-call-digest-v1
{"call_text":"<exact call text>","due_on":"YYYY-MM-DD","published_on":"YYYY-MM-DD"}
The file ends at the closing brace, with no trailing newline. Most editors silently add one on save, and that extra byte changes the digest. Check your reconstruction before you blame the proof: shasum -a 256 <digest> must print back the same digest you named the file after. If it does not, the file is wrong, not the timestamp.
Download the matching <digest>.ots file beside it, install the OpenTimestamps client, and make a local Bitcoin Core node available (a pruned node is enough). Then run:
ots verify <digest>.ots
A proof marked Pending is not verifiable yet. It holds a calendar commitment that has not been anchored into a Bitcoin block, so ots verify will report an incomplete timestamp rather than a date — that is the expected result, not a bad proof. Run ots upgrade <digest>.ots once the attestation lands, then verify.
How the record is kept
Read the operating rules behind the scoreboard, from publication through correction.
Keep exploring
Read the analysis, download the dataset, or inspect the evidence behind the record.