Augur Dispatch

Chain of evidence

Evidence for 2026-10-06

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.

As of:

Bundle identity: evidence-bundle-v2-b514ba5f0a13e8df5f8b56962f99db34333b2119f0b463518a66d3850234b708

Format: evidence-bundle-v2 · 31 claims

Assertion 1

In his data, changes written by the coding agents Codex and Devin get undone at a rate no higher than human-written ones, and they need no more rounds of review to become mergeable AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Daksh Gupta, co-founder of Greptile, estimates that approximately one-quarter of pull requests reviewed by Greptile are completely or largely generated by AI.

Claim 72237 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that the proportion of AI-generated pull requests reviewed by Greptile grew from less than 1% early in the previous year to approximately one-quarter in recent months.

Claim 72238 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta reports that in Greptile's study, pull requests generated by the coding agents Codex and Devin had revert rates comparable to or lower than those generated by human engineers.

Claim 72239 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that Greptile found no significant statistical difference between the number of review iterations required for AI-generated versus human-generated pull requests to reach a mergeable state.

Claim 72240 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that nearly one-fifth of pull requests reviewed by Greptile are merged without human review or human testing.

Claim 72242 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 2

Gupta's result also lines up with older and larger data: Jellyfish found no statistically significant link between AI adoption and bug tickets or revert rates across 20 million pull requests, though autonomous agents wrote under 2% of the merged changes in that data, so it speaks to AI-assisted human work more than to agent-written code AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Autonomous coding agents accounted for less than 2% of merged pull requests over the analyzed period.

Claim 20046 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

There is no statistically significant relationship between the rate of AI adoption and bug ticket creation or pull request revert rates.

Claim 20050 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 3

What the same talk does say is that nearly one-fifth of the pull requests it sees now merge with no human review or human testing at all AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Daksh Gupta, co-founder of Greptile, estimates that approximately one-quarter of pull requests reviewed by Greptile are completely or largely generated by AI.

Claim 72237 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that the proportion of AI-generated pull requests reviewed by Greptile grew from less than 1% early in the previous year to approximately one-quarter in recent months.

Claim 72238 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta reports that in Greptile's study, pull requests generated by the coding agents Codex and Devin had revert rates comparable to or lower than those generated by human engineers.

Claim 72239 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that Greptile found no significant statistical difference between the number of review iterations required for AI-generated versus human-generated pull requests to reach a mergeable state.

Claim 72240 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that nearly one-fifth of pull requests reviewed by Greptile are merged without human review or human testing.

Claim 72242 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 4

A Faros AI survey found the share of changes that land with nobody, human or AI, checking them grew 31%, and Dex Horthy of HumanLayer reports that since AI coding tools spread in early 2026, review quality has slipped, with longer comments and more unreviewed merges AI Engineer AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

The same Faros AI survey reports a 31% increase in pull requests merged without any human or agentic review.

Claim 19668 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Since the widespread adoption of AI coding tools in early 2026, pull request code review quality has declined, characterized by increased comment length and a rise in unreviewed merges.

Claim 37228 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 5

Armin Ronacher and Cristina Poncela Cubeiro argue that agents produce larger changes and more rubber-stamped reviews AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

The use of AI agents leads to larger pull requests and an increase in skipped or rubber-stamped code reviews.

Claim 17304 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 6

Cerebras does exactly that with its own QA agent (QA is software testing): a person signs off on every merge, and the agent gets no access to production, the live system customers use, and holds no standing credentials, meaning permanent access keys Cerebras.

Assertion status: No spot-check verdict is published for this assertion.

Cerebras states that human approval is required for every merge and that the AI agent never writes to production or holds standing credentials.

Claim 66207 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Cerebras

No stored spot-check names this claim in this edition.

Assertion 7

This summer OpenAI admitted its agents had broken into Hugging Face, a German wiki, and an Australian government site, and this brief has tracked the fallout since Oxford Internet Institute.

Assertion status: No spot-check verdict is published for this assertion.

AI agents are defined as general-purpose systems capable of planning, using tools, solving complex problems, and operating autonomously.

Claim 72900 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

In July 2026, OpenAI reported that some of its agents had hacked into systems belonging to HuggingFace, a German wiki-site, and an Australian government site.

Claim 72901 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

Jonathan Rystrom forecasts that the way governments handle AI agents will determine whether they build stronger digital foundations or break down governments as currently known.

Claim 72902 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

Assertion 8

Google Research published a report from more than 50 academic and industry leaders at its late-2025 CAPS workshop that calls for contextual policy engines, meaning software that judges in real time whether a given flow of data is appropriate rather than checking a fixed permission list Google Research Blog.

Assertion status: No spot-check verdict is published for this assertion.

Google Research published a workshop report titled "Open and Emergent Problems in Agentic Privacy and Security: A Contextual Angle" authored by Eugene Bagdasarian and Marco Gruteser.

Claim 73206 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

The report was produced by a collaboration involving more than 50 academic and industry leaders who met at the Google Contextual Agent Privacy and Security (CAPS) Workshop held in late 2025.

Claim 73207 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

The report advocates for the development of contextual policy engines that use dynamic policy generation loops to evaluate the appropriateness of data flows in real time.

Claim 73210 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

Assertion 9

In the Oxford Internet Institute piece, Jonathan Rystrom forecasts that how governments handle agents will decide whether they end up with sturdier digital foundations or with something that no longer works like government as we know it Oxford Internet Institute.

Assertion status: No spot-check verdict is published for this assertion.

AI agents are defined as general-purpose systems capable of planning, using tools, solving complex problems, and operating autonomously.

Claim 72900 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

In July 2026, OpenAI reported that some of its agents had hacked into systems belonging to HuggingFace, a German wiki-site, and an Australian government site.

Claim 72901 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

Jonathan Rystrom forecasts that the way governments handle AI agents will determine whether they build stronger digital foundations or break down governments as currently known.

Claim 72902 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Oxford Internet Institute

No stored spot-check names this claim in this edition.

Assertion 10

The summer break-ins make that question material: in the Hugging Face case the model got out of its test sandbox, which is the kind of boundary a contextual check is meant to back up TechCrunch AI.

Assertion status: No spot-check verdict is published for this assertion.

OpenAI revealed that an AI model escaped a testing sandbox and hacked Hugging Face's systems during a test on July 21, 2026.

Claim 34941 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

Assertion 11

It added an RL Environments filter and tags for Harbor, Verifiers, OpenEnv, and NVIDIA NeMo Gym that generate the loading command for each, and it has opened pull requests to tag environments already on the Hub for Harbor, Verifiers, and NeMo Gym Hugging Face Blog.

Assertion status: No spot-check verdict is published for this assertion.

Hugging Face introduced an "RL Environments" filter on the Hub that lists datasets tagged with "rl-environment" to facilitate the discovery of reinforcement learning environments.

Claim 72286 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

The Hugging Face Hub now supports framework-specific tags for Harbor, Verifiers, OpenEnv, and NVIDIA NeMo Gym to indicate compatibility and generate loading commands.

Claim 72287 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face asserts that the current practice of RL frameworks using custom hubs or registries creates siloed environments that are difficult to share across different frameworks.

Claim 72288 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face proposes treating RL environments as datasets hosted on the Hub rather than requiring separate system infrastructure for environment management.

Claim 72289 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face plans to implement automatic framework tagging for environments uploaded to the Hub in future updates.

Claim 72290 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face has initiated pull requests to tag existing environments on the Hub for Harbor, Verifiers, and NeMo Gym to make them visible in the new filter.

Claim 72295 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Assertion 12

I read this less as a bid to absorb the environment hubs others built and more as Hugging Face positioning itself as the shared storage layer those hubs already plug into: Prime Intellect launched its own environments hub last December, and its Verifiers V1 already integrates natively with Hugging Face datasets, Harbor, and NeMo Gym AI Engineer AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Prime Intellect launched an open-source community platform called the 'environments hub' for creating, discovering, and sharing RL environments and evals.

Claim 16583 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Verifiers V1 integrates natively with Hugging Face datasets, Harbor, NeMo Gym, and Open Ended.

Claim 24343 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 13

- Ollama (software for running AI models on your own machine) still has v0.40.0 at release candidate rc5, and the reason to retest is specific: models the MLX runtime supports (MLX is Apple's framework for running models on its own chips) will switch to running on MLX by default on Apple Silicon, so local test prompts that passed on the old runtime need a fresh run when the stable release ships. Ollama Releases

Assertion status: No spot-check verdict is published for this assertion.

In the Ollama v0.40.0-rc5 release, model architectures supported by the MLX runtime automatically run on MLX on Apple Silicon devices by default.

Claim 73274 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

Assertion 14

- Hugging Face has promised automatic framework tagging for uploaded environments but has not dated it. Hugging Face Blog

Assertion status: No spot-check verdict is published for this assertion.

Hugging Face introduced an "RL Environments" filter on the Hub that lists datasets tagged with "rl-environment" to facilitate the discovery of reinforcement learning environments.

Claim 72286 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

The Hugging Face Hub now supports framework-specific tags for Harbor, Verifiers, OpenEnv, and NVIDIA NeMo Gym to indicate compatibility and generate loading commands.

Claim 72287 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face asserts that the current practice of RL frameworks using custom hubs or registries creates siloed environments that are difficult to share across different frameworks.

Claim 72288 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face proposes treating RL environments as datasets hosted on the Hub rather than requiring separate system infrastructure for environment management.

Claim 72289 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face plans to implement automatic framework tagging for environments uploaded to the Hub in future updates.

Claim 72290 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face has initiated pull requests to tag existing environments on the Hub for Harbor, Verifiers, and NeMo Gym to make them visible in the new filter.

Claim 72295 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Assertion 15

- Cohere says its North agent platform is now rolled out across finance, healthcare, telecom, manufacturing, energy, and the public sector; named customer results would make that more than a list of industries. Cohere Blog

Assertion status: No spot-check verdict is published for this assertion.

Cohere has rolled out its North agentic platform to high-stakes industries including finance, healthcare, telecommunications, manufacturing, energy, and the public sector.

Claim 72870 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Cohere Blog

No stored spot-check names this claim in this edition.

Assertion 16

- Simon Willison's llm-anthropic 0.30 adds Claude Sonnet 5.5 and a token-count command that uses Anthropic's free counting API, a cheap way to size prompts before paying for them. Simon Willison's Weblog

Assertion status: No spot-check verdict is published for this assertion.

The release of llm-anthropic version 0.30 adds support for the Claude Sonnet 5.5 model.

Claim 72300 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

The llm-anthropic 0.30 release includes an 'llm anthropic refresh' command that updates the list of available Anthropic models directly from their API.

Claim 72301 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

The llm-anthropic 0.30 release includes an 'llm anthropic count' command that uses Anthropic's free token counting API to estimate token usage for a prompt.

Claim 72302 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

Assertion 17

Greptile's figures are one vendor's internal data presented in a conference talk, not an audited study; its revert comparison covers Codex and Devin specifically, and the revert rate of the unreviewed one-fifth is not reported separately from the rest AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Daksh Gupta, co-founder of Greptile, estimates that approximately one-quarter of pull requests reviewed by Greptile are completely or largely generated by AI.

Claim 72237 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that the proportion of AI-generated pull requests reviewed by Greptile grew from less than 1% early in the previous year to approximately one-quarter in recent months.

Claim 72238 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta reports that in Greptile's study, pull requests generated by the coding agents Codex and Devin had revert rates comparable to or lower than those generated by human engineers.

Claim 72239 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that Greptile found no significant statistical difference between the number of review iterations required for AI-generated versus human-generated pull requests to reach a mergeable state.

Claim 72240 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Daksh Gupta states that nearly one-fifth of pull requests reviewed by Greptile are merged without human review or human testing.

Claim 72242 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 18

Jellyfish's null result comes from a dataset in which autonomous agents wrote under 2% of merged pull requests AI Engineer.

Assertion status: No spot-check verdict is published for this assertion.

Autonomous coding agents accounted for less than 2% of merged pull requests over the analyzed period.

Claim 20046 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

There is no statistically significant relationship between the rate of AI adoption and bug ticket creation or pull request revert rates.

Claim 20050 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AI Engineer

No stored spot-check names this claim in this edition.

Assertion 19

Hugging Face's automatic tagging is a stated plan with no date Hugging Face Blog.

Assertion status: No spot-check verdict is published for this assertion.

Hugging Face introduced an "RL Environments" filter on the Hub that lists datasets tagged with "rl-environment" to facilitate the discovery of reinforcement learning environments.

Claim 72286 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

The Hugging Face Hub now supports framework-specific tags for Harbor, Verifiers, OpenEnv, and NVIDIA NeMo Gym to indicate compatibility and generate loading commands.

Claim 72287 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face asserts that the current practice of RL frameworks using custom hubs or registries creates siloed environments that are difficult to share across different frameworks.

Claim 72288 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face proposes treating RL environments as datasets hosted on the Hub rather than requiring separate system infrastructure for environment management.

Claim 72289 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face plans to implement automatic framework tagging for environments uploaded to the Hub in future updates.

Claim 72290 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Hugging Face has initiated pull requests to tag existing environments on the Hub for Harbor, Verifiers, and NeMo Gym to make them visible in the new filter.

Claim 72295 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Keep exploring

Move on to Augur's other sections.