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 →
96
264
95
0
5
Recent pipeline events
| When | Event | Piece | Detail |
|---|---|---|---|
| 31 Oct, 16:35 | ∅ held back | nemotron-bolts-benchmark-edge-meets-a-trust-problem | critic unavailable |
| 31 Oct, 12:38 | ∅ held back | fast-model-swapping-needs-a-timing-test | critic unavailable |
| 31 Oct, 07:07 | ∅ held back | deepseek-updates-v4-1-flash-after-outage | critic unavailable |
| 30 Oct, 16:37 | ∅ held back | google-opens-android-coding-tools-to-rival-agents | critic unavailable |
| 30 Oct, 12:34 | ∅ held back | aws-publishes-a-blueprint-for-governed-ai-agents | critic unavailable |
| 30 Oct, 07:13 | ∅ held back | aws-publishes-a-guide-to-visual-ai-agents | critic unavailable |
| 29 Oct, 16:40 | ✎ drafted | Prism ships a runnable three-value AI model | evening slot |
| 29 Oct, 12:37 | ∅ held back | openbonsai-compresses-custom-ai-models | critic unavailable |
| 29 Oct, 07:09 | ∅ held back | google-releases-gemini-4-carbon-to-cyber-defenders | critic unavailable |
| 28 Oct, 16:40 | ∅ held back | legora-scales-into-the-legal-operating-system | critic unavailable |
| 28 Oct, 12:46 | ✎ drafted | Legora launches an AI system for legal work | lunch slot |
| 28 Oct, 07:09 | ∅ held back | parallel-ai-agents-buy-speed-not-efficiency | critic unavailable |
| 27 Oct, 16:42 | ∅ held back | the-right-local-ai-card-depends | critic unavailable |
| 27 Oct, 12:43 | ∅ held back | qwen-3-5-9b-starts-work-then-stops | critic unavailable |
| 27 Oct, 07:10 | ∅ held back | six-ai-engines-tested-on-one-graphics-card | critic unavailable |
| 26 Oct, 16:47 | ∅ held back | ai-agents-can-now-organise-their-own-teams | critic unavailable |
| 26 Oct, 12:40 | ∅ held back | run-qwen3-8-flash-next-on-strix-halo | critic unavailable |
| 26 Oct, 07:12 | ∅ held back | local-ai-training-avoids-a-low-memory-shortcut | critic unavailable |
| 25 Oct, 16:40 | ∅ held back | nvidia-releases-flora-for-robot-video-labelling | critic unavailable |
| 25 Oct, 12:44 | ∅ held back | legacy-nvidia-gpus-get-a-community-compatibility-patch | critic unavailable |
| 25 Oct, 07:09 | ∅ held back | minimaxs-new-ai-model-gets-day-one-local | critic unavailable |
| 24 Oct, 16:38 | ∅ held back | swarm-organises-parallel-coding-agents | critic unavailable |
| 24 Oct, 12:39 | ✎ drafted | Qwen 3.8 nearly ties Claude on Korean PDFs | lunch slot |
| 24 Oct, 07:10 | ∅ held back | nvidia-pitches-europe-as-an-ai-production-hub | critic unavailable |
| 23 Oct, 16:43 | ∅ held back | google-isolates-every-gemini-agent | critic unavailable |
| 23 Oct, 12:44 | ✎ drafted | Reusable AI instructions get a neutral home | lunch slot |
| 23 Oct, 07:13 | ✎ drafted | Muse Spark benchmark favours TensorSharp | morning slot |
| 22 Oct, 16:37 | ∅ held back | recursive-releases-flint-for-macbooks | critic unavailable |
| 22 Oct, 12:44 | ✎ drafted | Anthropic previews a background assistant called Duet | lunch slot |
| 22 Oct, 07:17 | ✎ drafted | AWS shows how to route AI work locally | morning slot |
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.