Transparency
Inside the newsroom
This publication is a working experiment: an AI agent finds the stories, researches and writes them; automated gates reject anything it can't prove from its sources; and a human editor approves every article before it goes live. This page is the live record of that pipeline — including the drafts that didn't make it. How we built it →
98
271
126
0
5
Recent pipeline events
| When | Event | Piece | Detail |
|---|---|---|---|
| 13 Nov, 16:38 | ∅ held back | nvidia-blackwell-fingerprints-found-in-six-chinese-models | critic unavailable |
| 13 Nov, 12:38 | ∅ held back | gemini-4-hydrogen-beats-argon-in-four-weeks | critic unavailable |
| 13 Nov, 07:20 | ✎ drafted | Model choice grows as NVIDIA faces scrutiny | morning slot |
| 12 Nov, 16:39 | ∅ held back | ai-now-splits-into-specialist-tool-markets | critic unavailable |
| 12 Nov, 12:38 | ∅ held back | claude-fable-5-2-targets-demanding-technical-work | critic unavailable |
| 12 Nov, 07:09 | ∅ held back | ai-model-costs-fall-as-nvidia-scrutiny-rises | critic unavailable |
| 11 Nov, 16:43 | ∅ held back | model-releases-are-splitting-into-price-bands | critic unavailable |
| 11 Nov, 12:38 | ∅ held back | a-practical-test-for-metas-muse-ember | critic unavailable |
| 11 Nov, 07:11 | ∅ held back | google-opens-gemini-4-carbon-for-paid-use | critic unavailable |
| 10 Nov, 16:41 | ✎ drafted | AI's benchmark leader has a steep price | evening slot |
| 10 Nov, 12:48 | ∅ held back | fable-5-2-turns-a-tweet-into-video | staging verification failed after 8 min |
| 10 Nov, 07:23 | ∅ held back | two-ai-models-arrive-for-demanding-technical-work | critic unavailable |
| 09 Nov, 16:35 | ∅ held back | open-and-closed-models-leave-bioweapon-risk-exposed | critic unavailable |
| 09 Nov, 12:34 | ∅ held back | test-whether-memory-is-slowing-your-ai | critic unavailable |
| 09 Nov, 07:09 | ∅ held back | korea-adds-controls-for-ai-agents | critic unavailable |
| 08 Nov, 16:38 | ∅ held back | crowd-v1-is-a-model-trained-by-volunteers | critic unavailable |
| 08 Nov, 12:39 | ∅ held back | a-panel-of-models-beats-a-single-judge | critic unavailable |
| 08 Nov, 07:11 | ✎ drafted | Open model targets bot interruptions in multilingual calls | morning slot |
| 07 Nov, 16:34 | ∅ held back | a-tiny-model-changes-ai-document-search | critic unavailable |
| 07 Nov, 12:43 | ∅ held back | ripple-65b-preview-opens-for-coding-tests | critic unavailable |
| 07 Nov, 07:08 | ∅ held back | unsloth-refreshes-its-embedding-model-collection | critic unavailable |
| 06 Nov, 16:38 | ∅ held back | ai-agents-score-just-12-on-work-benchmark | critic unavailable |
| 06 Nov, 12:41 | ✎ drafted | Open-model security skills need a controlled test | lunch slot |
| 06 Nov, 07:05 | ∅ held back | microsoft-puts-agentic-coding-on-device | critic unavailable |
| 05 Nov, 16:45 | ✎ drafted | Open-source PipesHub tops the enterprise search benchmark | evening slot |
| 05 Nov, 12:41 | ∅ held back | unsloth-releases-new-qwen3-8-flash-next-files | critic unavailable |
| 05 Nov, 07:19 | ✎ drafted | DeepSeek V4.1 Flash draws a bullish verdict | morning slot |
| 04 Nov, 16:36 | ∅ held back | tools-change-what-small-agents-can-do | critic unavailable |
| 04 Nov, 12:41 | ∅ held back | turn-an-agent-benchmark-into-an-acceptance-test | critic unavailable |
| 04 Nov, 07:11 | ∅ held back | smallest-model-passes-63-8-of-agent-tasks | critic unavailable |
Why publish the failures? Because "an AI wrote this" only deserves trust if you can see the times it wasn't good enough — and who decided. The gates and the editor say no more often than the marketing around AI suggests they should.