This frozen page shows Augur's claims and source links for one sent dispatch. Stored spot-checks appear only where the frozen edition supports them; absence is not presented as verification.
The early evidence that fed that view measured groups, not individuals: across 758 Boston Consulting Group consultants, access to AI lifted performance on routine assignments, and a related BCG and Harvard analysis put the gain at 12% more tasks finished 25% faster when the work sat inside what the model could handle One Useful Thing Nate Jones.
Assertion status: No spot-check verdict is published for this assertion.
A study of 758 Boston Consulting Group consultants found that those using AI vastly outperformed those without on standard tasks.
Claim 1105Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Consultants working on tasks outside the AI's capability frontier were 19 percentage points less likely to achieve correctness compared to those working without AI. fact null null
Claim 14119Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 2
The juniors are the turn in the story: as a group they ended no more skilled than before, because the spread widened, with more of them scoring high and more scoring low Google Research Blog.
Assertion status: No spot-check verdict is published for this assertion.
In a three-month randomized controlled trial involving 133 patent attorneys, routine use of an AI patent writing assistant improved work quality across all participants. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74835Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Senior lawyers who used the AI assistant for 90 days demonstrated stronger professional judgment than senior lawyers who did not use the tool. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74836Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Junior lawyers showed no average skill gain from using the AI assistant; their performance scores bifurcated into more high scores and more low scores. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74837Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 3
The older consulting data offers a warning sign, though: when consultants hit a problem the model could not handle, those with AI were 19 percentage points less likely to get it right than those working alone One Useful Thing Nate Jones.
Assertion status: No spot-check verdict is published for this assertion.
A study of 758 Boston Consulting Group consultants found that those using AI vastly outperformed those without on standard tasks.
Claim 1105Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Consultants working on tasks outside the AI's capability frontier were 19 percentage points less likely to achieve correctness compared to those working without AI. fact null null
Claim 14119Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 4
Data from more than 400 software organizations points the same way: junior engineers use the tools more often, while staff-level engineers, the most senior hands-on engineers, save about as much time on fewer tokens, the chunks of text a model reads and writes and the unit vendors bill by AI Engineer.
Assertion status: No spot-check verdict is published for this assertion.
Junior engineers are using AI tools more frequently than senior engineers, but staff-plus level engineers save a similar amount of time while using fewer tokens.
Claim 72323Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 5
The open decision is framing, and Stanford's argument is that staff who are sold generative AI as a way to get more done tend to leave it alone Stanford HAI.
Assertion status: No spot-check verdict is published for this assertion.
In 2023, 55% of organizations worldwide reported using AI.
Claim 74475Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 6
Anthropic said on September 28 that Haiku 5.5 would arrive within weeks, and until yesterday its cheapest model was Haiku 4.5 at $1 per million input tokens and $5 per million output tokens Simon Willison's Weblog Simon Willison's Weblog.
Assertion status: No spot-check verdict is published for this assertion.
Anthropic's cheapest current model is Claude Haiku 4.5, with prices of $1 for input and $5 for output per million tokens.
Claim 40201Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 7
Haiku 5.5 is now out, and Anthropic puts the average saving at roughly 75% against Haiku 4.5, yet the average hides a split: requests up to 100,000 tokens are priced 90% lower, requests above that only 50% lower Anthropic News.
Assertion status: No spot-check verdict is published for this assertion.
Anthropic released Claude Haiku 5.5, which it describes as its cheapest, fastest, and most capable small model to date.
Claim 74414Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 8
NVIDIA ran the IOI test as it happened, under the same rules the students faced, but that does not make it official: nobody supervised the run, and the score sits outside the contest's real ranking Hugging Face Blog.
Assertion status: No spot-check verdict is published for this assertion.
The Nemotron-3-Ultra-CC model, fine-tuned with supervised fine-tuning and the GenCorrect inference strategy, scored 535.4 out of 600 on the IOI 2026 competition, exceeding the gold medal threshold of 361.12 and the top human score of 498.27.
Claim 74500Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
The Nemotron 3 Ultra model, utilizing supervised fine-tuning and reinforcement learning checkpoints within a generate-verify-refine system, scored 30 out of 42 on the IMO 2026 competition, exceeding the official gold medal threshold of 29.
Claim 74501Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
The IOI 2026 result was generated via a live, prospective run under constraints identical to those for human contestants, but it remains an unofficial, unsupervised benchmark and is not part of the official IOI ranking.
Claim 74502Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 9
Lab-run olympiad claims generally sit outside official judging, and an earlier OpenAI IMO gold claim was made without the official private marking guide Nate Jones.
Assertion status: No spot-check verdict is published for this assertion.
OpenAI's claim of winning an International Math Olympiad gold medal was made without access to the official private marking guide, creating uncertainty about whether the score met the strict criteria for the award.
Claim 12168Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 11
- Anthropic's tiered Haiku 5.5 pricing will meet real traffic soon, and the thing to track is the 100,000-token break: whether workloads built on long requests see only the smaller 50% cut in practice while short-request traffic gets the full 90%. Anthropic News
Assertion status: No spot-check verdict is published for this assertion.
Anthropic released Claude Haiku 5.5, which it describes as its cheapest, fastest, and most capable small model to date.
Claim 74414Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 12
- Google has kept Gemini 4 Argon inside a limited group of cyber defenders and says it will expand access; the move to paying API customers is the next checkpoint. Nate Jones
Assertion status: No spot-check verdict is published for this assertion.
Google's Gemini 4 Argon model has achieved impressive benchmark results on the Vals Index, which evaluates professional work across several fields.
Claim 74508Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 13
- Microsoft Research Asia's Agent Lightning v1.0 puts the same agent harness used in deployment, the software wrapper that runs an agent in production, directly inside reinforcement learning training, which is training by trial and reward, in about 3,500 lines of code; follow-on results on production tasks will show whether that approach holds beyond lab benchmarks. Microsoft Research Blog
Assertion status: No spot-check verdict is published for this assertion.
Microsoft Research Asia introduced a training paradigm called Harnessed Agentic RL, in which the same agent harness used in deployment participates directly in reinforcement learning.
Claim 74700Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 14
- OpenAI's global rollout of GPT-6 inside ChatGPT brings faster responses and an Intelligent UI with visuals and interactive experiences; the question for IT leaders is whether company accounts get the same behavior as consumer ones. OpenAI News
Assertion status: No spot-check verdict is published for this assertion.
OpenAI has announced the global rollout of GPT-6 within ChatGPT, featuring an Intelligent UI that provides faster responses along with visuals and interactive experiences.
Claim 74434Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 15
- Ai2's Bolmo work, now published in Nature, comes with new checkpoints that extend its byte-level approach, in which a model reads raw characters instead of word chunks, beyond the Olmo family, and uptake by other model builders would signal whether byte-level models move out of research. Allen Institute for AI
Assertion status: No spot-check verdict is published for this assertion.
Allen Institute for AI published research on Bolmo, a family of fully open byte-level language models, in the journal Nature.
Claim 74489Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Allen Institute released new checkpoints on Hugging Face that extend Bolmo's byteifying approach to other model families beyond its starting point, the Olmo models.
Claim 74490Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 16
The patent attorney findings come from the authors' Google Research blog post; the evidence does not say whether the study is peer-reviewed, and the sample is 133 lawyers in one profession, so the junior split is directional until replicated Google Research Blog.
Assertion status: No spot-check verdict is published for this assertion.
In a three-month randomized controlled trial involving 133 patent attorneys, routine use of an AI patent writing assistant improved work quality across all participants. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74835Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Senior lawyers who used the AI assistant for 90 days demonstrated stronger professional judgment than senior lawyers who did not use the tool. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74836Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Junior lawyers showed no average skill gain from using the AI assistant; their performance scores bifurcated into more high scores and more low scores. (Source: David Autor and Tanya Rodchenko, Google Research, 2026-10-07)
Claim 74837Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
Assertion 17
The NVIDIA IOI score is an unofficial run outside the official ranking, and the evidence does not state the IMO run's official status beyond its clearing the official gold threshold Hugging Face Blog.
Assertion status: No spot-check verdict is published for this assertion.
The Nemotron-3-Ultra-CC model, fine-tuned with supervised fine-tuning and the GenCorrect inference strategy, scored 535.4 out of 600 on the IOI 2026 competition, exceeding the gold medal threshold of 361.12 and the top human score of 498.27.
Claim 74500Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
The Nemotron 3 Ultra model, utilizing supervised fine-tuning and reinforcement learning checkpoints within a generate-verify-refine system, scored 30 out of 42 on the IMO 2026 competition, exceeding the official gold medal threshold of 29.
Claim 74501Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded
No stored spot-check names this claim in this edition.
The IOI 2026 result was generated via a live, prospective run under constraints identical to those for human contestants, but it remains an unofficial, unsupervised benchmark and is not part of the official IOI ranking.
Claim 74502Label: factProvenance: primaryEligible for corroboration · eligibility does not establish truthRecorded