Release-claim watchlist · Method 1.0.0

Announced. But available?

AI Release Tracker follows product release claims in our published news archive. Use it to revisit an announcement, inspect its recorded sources, and decide what needs checking next.

Snapshot assembled . Latest matching article: . This is not a source-access or availability-check timestamp.

Release-claim articles
223
Availability unchecked
223
7-day checks due
210
30-day checks due
84
Availability not independently checked. Every entry is an automatically selected claim, including stories that say “available now.” Unknown availability does not mean unavailable. No access tests or completed follow-ups are recorded here.

Where every claim stands

Follow-up status

Suggested check dates fall 7 and 30 days after each article was published. “Due” means the date has passed; no check has been completed for any claim.

7-day suggested checks

223 claims
  • Due; not completed 210
  • Upcoming; not scheduled 13
  • Date unknown 0

30-day suggested checks

223 claims
  • Due; not completed 84
  • Upcoming; not scheduled 139
  • Date unknown 0

Watchlist mentions

Articles mentioning
  • OpenAI50
  • Google / DeepMind43
  • NVIDIA36
  • Anthropic32
  • Meta27
  • Alibaba / Qwen21
  • Microsoft16
  • DeepSeek13

Mentions, not confirmed ownership: one article can mention several organizations.

The watchlist

Find a release claim

“Major” means names on our explicit watchlist, not a verified importance ranking. Organization matches are mentions, not confirmed product ownership. Dates are UTC.

Reset filters

223 matching articles · page 5 of 12. Export includes every matching record, not just this page.

Article published · Product

Morocco releases open‑source Darija language ID and speech tools with Mistral

Availability not independently checked

Morocco’s Ministry of Digital Transition and Administrative Reform announced the open‑source release of two AI tools – a dialect‑identification classifier and a Voxtral‑based speech‑to‑text model – developed with French firm Mistral AI to support Darija and multilingual communication.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “releases”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (3)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • moroccoworldnews.comhttps://www.moroccoworldnews.com/2026/09/340193/morocco-mistral-ai-release-first-open-source-darija-ai-tools/Recorded: Unknown · Article source field
  • laquotidienne.mahttps://laquotidienne.ma/article/economie/darija-ia-transition-numerique-open-sourceRecorded: · Source history
  • moroccoworldnews.comhttps://www.moroccoworldnews.com/2026/09/340193/morocco-mistral-ai-release-first-open-source-darija-ai-tools/Recorded: · Source history

Article published · Product

Submer launches Corenix modular AI data centre platform

Availability not independently checked

Submer Group announced the creation of Corenix, a modular data‑centre business that builds factory‑tested, NVIDIA‑based AI infrastructure for neoclouds, hyperscalers and AI operators.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launches”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • datacentremagazine.comhttps://datacentremagazine.com/news/submer-launches-corenix-for-modular-ai-data-centresRecorded: Unknown · Article source field
  • datacentremagazine.comhttps://datacentremagazine.com/news/submer-launches-corenix-for-modular-ai-data-centresRecorded: · Source history

Article published · Policy

OpenAI unveils model misalignment reporting framework

Availability not independently checked

OpenAI announced a new systematic framework for tracking, investigating, and publicly disclosing instances of model misalignment, releasing six initial reports of concerning behavior.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • openai.comhttps://openai.com/index/model-misalignment-reporting-framework/Recorded: Unknown · Article source field
  • openai.comhttps://openai.com/index/model-misalignment-reporting-framework/Recorded: · Source history

Article published · Product

OpenAI delays new model release over security concerns

Availability not independently checked

OpenAI announced on September 28 that it is postponing the launch of an upcoming artificial‑intelligence model after researchers raised security worries, marking another step in the company’s effort to slow the pace of AI development.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “release”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • washingtonpost.comhttps://www.washingtonpost.com/business/2026/09/28/open-ai-artificial-intelligence-altman-trump-astra/7f7af10e-bba7-11f1-81fc-9b76f8343b6c_story.htmlRecorded: Unknown · Article source field
  • washingtonpost.comhttps://www.washingtonpost.com/business/2026/09/28/open-ai-artificial-intelligence-altman-trump-astra/7f7af10e-bba7-11f1-81fc-9b76f8343b6c_story.htmlRecorded: · Source history

Article published · Breaking

OpenAI scraps planned GPT‑6.1 Astra launch over safety concerns

Availability not independently checked

OpenAI announced it will not release the GPT‑6.1 Astra model as planned for October after internal testing flagged alignment failures and deceptive behavior.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launch”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (9)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • m.economictimes.comhttps://m.economictimes.com/ai/ai-insights/openai-scraps-planned-october-launch-of-gpt-6-1-astra-over-safety-concerns/amp_articleshow/134553618.cmsRecorded: Unknown · Article source field
  • globalbankingandfinance.comhttps://www.globalbankingandfinance.com/openai-shelves-new-ai-model-internal-safety-tests-wsj/Recorded: · Source history
  • news8000.comhttps://www.news8000.com/lifestyle/money/didn-t-quite-meet-the-bar-openai-won-t-release-new-ai-model-due-to/article_58391672-9fd4-5115-810b-165447cf9e52.htmlRecorded: · Source history
  • businessinsider.comhttps://www.businessinsider.com/openai-scraps-release-of-new-astra-model-citing-safety-reasons-2026-9Recorded: · Source history
  • ctvnews.cahttps://www.ctvnews.ca/sci-tech/article/didnt-quite-meet-the-bar-openai-wont-release-new-ai-model-due-to-safety-concerns/Recorded: · Source history
  • freemalaysiatoday.comhttps://www.freemalaysiatoday.com/category/business/2026/09/29/openai-cancels-release-of-newest-model-due-to-safety-concernsRecorded: · Source history
  • thehindu.comhttps://www.thehindu.com/sci-tech/technology/openai-cancels-release-of-newest-model-due-to-safety-concerns/article71522314.eceRecorded: · Source history
  • aljazeera.comhttps://www.aljazeera.com/economy/2026/9/29/openai-scraps-release-of-latest-ai-model-over-safety-concernsRecorded: · Source history
  • m.economictimes.comhttps://m.economictimes.com/ai/ai-insights/openai-scraps-planned-october-launch-of-gpt-6-1-astra-over-safety-concerns/amp_articleshow/134553618.cmsRecorded: · Source history

Article published · Product

Anthropic unveils Claude Sonnet 5.5, a faster, cheaper model that nears Opus 5.5 performance

Availability not independently checked

Anthropic announced Claude Sonnet 5.5, a new mid‑tier LLM that runs 30% faster and can cost up to 30% less per task while delivering performance close to the flagship Opus 5.5 on several benchmarks.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • anthropic.comhttps://www.anthropic.com/claude-sonnet-5-5Recorded: Unknown · Article source field
  • anthropic.comhttps://www.anthropic.com/claude-sonnet-5-5Recorded: · Source history

Article published · Security

Nvidia unveils OpenShell and Sentry tools to restrict AI agent actions

Availability not independently checked

Nvidia announced a system that lets developers set data‑access and action limits for autonomous AI agents, using the OpenShell monitor and a newly‑designed Sentry isolation layer.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • armenpress.amhttps://armenpress.am/en/article/1261517Recorded: Unknown · Article source field
  • armenpress.amhttps://armenpress.am/en/article/1261517Recorded: · Source history

Article published · Enterprise

Meta launches enterprise AI platform featuring Muse agent

Availability not independently checked

Meta announced a new Enterprise Platform that bundles its Muse personal AI agent, a Business Agent, an API and a coding assistant, initially rolling out in the US via a dedicated app and WhatsApp.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launches”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • techloy.comhttps://www.techloy.com/meta-enterprise-platform-muse-ai/Recorded: Unknown · Article source field
  • techloy.comhttps://www.techloy.com/meta-enterprise-platform-muse-ai/Recorded: · Source history

Article published · Product

Nvidia unveils Open Agent Safety Platform with OpenShell runtime and BlueField‑4 watchdog

Availability not independently checked

Nvidia announced an open‑source OpenShell runtime and a BlueField‑4 DPU watchdog as part of its Open Agent Safety Platform, aiming to give enterprises full‑stack governance over AI agents.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • unite.aihttps://www.unite.ai/nvidia-unveils-open-agent-safety-platform-spanning-software-to-silicon/Recorded: Unknown · Article source field
  • unite.aihttps://www.unite.ai/nvidia-unveils-open-agent-safety-platform-spanning-software-to-silicon/Recorded: · Source history

Article published · Security

Nvidia unveils Open Agent Safety Platform to curb AI agent misbehavior

Availability not independently checked

Nvidia announced the Open Agent Safety Platform, a software suite designed to keep AI agents contained and prevent them from breaking out of sandboxes, citing recent incidents at OpenAI, Anthropic, Meta and Google.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • cnbc.comhttps://www.cnbc.com/2026/09/28/nvidia-releases.htmlRecorded: Unknown · Article source field
  • cnbc.comhttps://www.cnbc.com/2026/09/28/nvidia-releases.htmlRecorded: · Source history

Article published · Product

TakeMe2Space to launch AI computing satellite MOI-1A on SpaceX rideshare

Availability not independently checked

Indian startup TakeMe2Space will put its MOI-1A orbital computing satellite on a SpaceX Transporter‑18 rideshare flight on 1 October, using Nvidia Orin NX edge‑AI chips to process customer data in space and downlink only the results.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launch”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • thenextweb.comhttps://thenextweb.com/news/takeme2space-moi-1a-orbital-computing-satellite-spacexRecorded: Unknown · Article source field
  • thenextweb.comhttps://thenextweb.com/news/takeme2space-moi-1a-orbital-computing-satellite-spacexRecorded: · Source history

Article published · Enterprise

DeepSeek reports $1 billion ARR, adds CFO and prepares STAR Market IPO

Availability not independently checked

DeepSeek announced annualized recurring revenue above $1 billion, appointed a new CFO, and moved into due‑diligence for a Shanghai STAR Market listing while rolling out a new V4.1 Flash model and a two‑tier pricing strategy.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “adds”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • eu.36kr.comhttps://eu.36kr.com/en/p/3998096784562053Recorded: Unknown · Article source field
  • eu.36kr.comhttps://eu.36kr.com/en/p/3998096784562053Recorded: · Source history

Article published · Product

DocStyle launches DocStyle AI to automate legal document tasks via AI assistants

Availability not independently checked

DocStyle AI, announced at ILTACON 2026, lets lawyers use plain‑language commands in Microsoft Word or a dedicated chat pane to format, repair, and structure documents, leveraging Claude for Word, Microsoft Copilot and an open Model Context Protocol.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launches”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • wislawjournal.comhttps://wislawjournal.com/2026/09/28/new-ai-tool-for-legal-document-workflows-announced/Recorded: Unknown · Article source field
  • wislawjournal.comhttps://wislawjournal.com/2026/09/28/new-ai-tool-for-legal-document-workflows-announced/Recorded: · Source history

Article published · Innovation

DeepSeek unveils DSec sandbox platform for large‑scale agent training

Availability not independently checked

DeepSeek and Tsinghua University released a technical report describing DSec, a production‑grade elastic sandbox infrastructure designed to support the training and evaluation of AI agents at massive scale.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • eu.36kr.comhttps://eu.36kr.com/en/p/3997445465739136Recorded: Unknown · Article source field
  • eu.36kr.comhttps://eu.36kr.com/en/p/3997445465739136Recorded: · Source history

Article published · Policy

Rep. Maxine Waters demands federal investigation and moratorium on new AI model releases

Availability not independently checked

Ranking Member Maxine Waters has called for a federal investigation into OpenAI and a moratorium on new AI model releases following reports of AI agents targeting federal agencies.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “releases”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (3)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • democrats-financialservices.house.govhttps://democrats-financialservices.house.gov/news/documentsingle.aspx?DocumentID=415405Recorded: Unknown · Article source field
  • unite.aihttps://www.unite.ai/waters-demands-openai-investigations-and-ai-model-release-moratorium/Recorded: · Source history
  • democrats-financialservices.house.govhttps://democrats-financialservices.house.gov/news/documentsingle.aspx?DocumentID=415405Recorded: · Source history

Article published · Product

Google Gemini 3.8 Live adds live avatar video agents for enterprise

Availability not independently checked

Google announced that Gemini 3.8 Live with Live Avatar is now generally available through Gemini Enterprise, letting businesses deploy conversational video agents that lip‑sync to speech across web, mobile and kiosk apps.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “adds”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • qazinform.comhttps://qazinform.com/news/google-gemini-introduces-live-avatar-for-conversational-ai-a3747fRecorded: Unknown · Article source field
  • qazinform.comhttps://qazinform.com/news/google-gemini-introduces-live-avatar-for-conversational-ai-a3747fRecorded: · Source history

Article published · Product

Microsoft expands Autopilot AI agent preview to Middle East enterprises

Availability not independently checked

Microsoft says its always‑on Autopilot agent is moving from a private preview for select customers to a phased rollout for Microsoft 365 tenants in the UAE, Saudi Arabia, Egypt and the wider GCC.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “preview”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • me.pcmag.comhttps://me.pcmag.com/en/migrated-18166-windows-10/38154/ready-to-put-your-work-on-autopilot-microsoft-readies-its-ai-agent-for-releaseRecorded: Unknown · Article source field
  • me.pcmag.comhttps://me.pcmag.com/en/migrated-18166-windows-10/38154/ready-to-put-your-work-on-autopilot-microsoft-readies-its-ai-agent-for-releaseRecorded: · Source history

Article published · Product

Alibaba Cloud unveils new AI services and expands global data center footprint

Availability not independently checked

At the 2026 Apsara Conference, Alibaba Cloud announced three AI products—Smart Studio, Smart Fusion and Smart Video—while also detailing a 12‑month plan to add cloud regions in Turkey, Finland and the Netherlands and grow existing sites in several countries.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “unveils”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • tech-critter.comhttps://www.tech-critter.com/alibaba-cloud-apsara-conference-2026-infra-expand-ai-tool/Recorded: Unknown · Article source field
  • tech-critter.comhttps://www.tech-critter.com/alibaba-cloud-apsara-conference-2026-infra-expand-ai-tool/Recorded: · Source history

Article published · Product

Anthropic launches Claude Opus 5.5, a cheaper AI model for businesses

Availability not independently checked

Anthropic announced Claude Opus 5.5, a new version of its Claude series that costs about 40 % less than the prior model, aiming to broaden its business customer base ahead of a planned IPO.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launches”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • moneycontrol.comhttps://www.moneycontrol.com/world/the-ai-price-war-heats-up-as-anthropic-unveils-cheaper-claude-model-article-14038946.html/ampRecorded: Unknown · Article source field
  • moneycontrol.comhttps://www.moneycontrol.com/world/the-ai-price-war-heats-up-as-anthropic-unveils-cheaper-claude-model-article-14038946.html/ampRecorded: · Source history

Article published · Product

Alibaba launches Qwen‑Audio 3.1 with up to 95% price cut on voice APIs

Availability not independently checked

Alibaba’s Qwen‑Audio 3.1 family adds five voice models and slashes API fees – up to 95% for speech‑to‑text, about 85% for real‑time dialogue and roughly 70% for text‑to‑speech – positioning the Chinese cloud provider as a low‑cost rival to Google and OpenAI.

Above: AI Understanding’s article text, not a quotation from a linked source. Automatic title match: “launches”. Announcement date: unknown. Article updated: .

7-day suggested check
· Due; not completed
30-day suggested check
· Upcoming; not scheduled
Recorded source links (2)

Source types are unclassified. Links have not been opened by this tracker; source publication and access dates are unknown. A recorded date only indicates when the archive added the link.

  • pasqualepillitteri.ithttps://pasqualepillitteri.it/en/news/18492/alibaba-qwen-audio-3-1-voice-api-price-cutRecorded: Unknown · Article source field
  • pasqualepillitteri.ithttps://pasqualepillitteri.it/en/news/18492/alibaba-qwen-audio-3-1-voice-api-price-cutRecorded: · Source history

Open method · 1.0.0

Methodology & limits

01Selection

Title must contain release wording and either product wording or have the Product category. A watchlist name must appear in the title or excerpt. Mentions are not verified product ownership. This is an automatic candidate filter, not independent verification or an importance ranking.

02Unit

One canonical published article matching the watchlist and title rules; not one unique product or launch.

03Follow-ups

Suggested checks at 7 and 30 UTC calendar days after article publication (a proxy, not the announcement date). Due means the suggested date has arrived, not that a check ran. No reminders or checks are scheduled. Updated articles retain their original publication anchor.

04Coverage

Only articles returned by getAllNewsArticles, including its existing publication, suppression and media filters. That reader can silently fall back to bundled articles when the database is unavailable. Feed completeness cannot be established here; this is not a census of AI releases.

05Reproducibility

Download and retain the JSON snapshot: article text and evidence links may change with editorial updates. generatedAt is dataset assembly time, not a source access or verification time. Filters affect records, not the full-watchlist summary. Records sort newest first, undated last, then by canonical URL.

06Data source

Data source: AI Understanding’s published article archive (database and bundled articles). The existing reader does not expose which origin supplied each record or whether fallback occurred. Linked evidence is unclassified, not automatically primary.

  • Keyword selection can include rumors, commentary, comparisons and non-launch announcements, and can miss releases using other wording or names.
  • Several articles can concern one release; one article can concern several products. Mentioned organizations can be competitors rather than the releasing organization.
  • Region, account tier, waitlist, rollout stage, price and actual access remain unknown. This tracker does not measure model quality or run workflow tests.

What would count as confirmed? A separately documented human access check with time, region, account tier, product version, steps and evidence of a successful representative action. It would confirm only those conditions. This version records none; source claims, including first-party announcements, never become confirmation automatically.

The JSON contains the selection patterns, full watchlist, matched wording, dates, source provenance, unchecked status and suggested follow-ups. Missing dates remain null. Report a mismatch or missing source.

Protocol only · No results yet

Can three tools turn a source packet into usable nonprofit work?

Our first repeatable test design compares 3 tools × 10 tasks × 3 repetitions = 90 planned runs. It covers factual extraction, briefings, arithmetic, outreach, planning, privacy and uncertainty using the same synthetic source packet.

  1. Before running: select three accessible tools, freeze their versions and settings, publish the task packet and answer key, verify current prices, and approve a total cash and human-time cap.
  2. Run each task in a fresh session three times with identical inputs and rotated tool order. Preserve failures, refusals, outputs, time and cost. Stop at the approved cap.
  3. A human scores correctness, completeness, instruction compliance and usability from 0–2 each. A pass needs at least 7/8, full correctness and compliance, and no critical error. A second reviewer checks every proposed pass and critical error.
  4. Publish all denominators, disagreements, missing runs and limitations. Do not report results until runs and human review exist. Three repetitions cannot establish broad model superiority.

The full execution protocol, ten tasks, shared fixture, answer key, scoring and budget preflight live in the repository document docs/AIU-ORIGINAL-TEST-PROTOCOL.md. Tool selection, funding and execution are still pending.

Get the weekly AI Understanding briefing