LIVE
Architectural tweaks may break conventional scaling law exponents16/09/26|OpenAI publishes a framework for reporting model misalignment16/09/26 · OpenAI|NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Architectural tweaks may break conventional scaling law exponents16/09/26|OpenAI publishes a framework for reporting model misalignment16/09/26 · OpenAI|NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|
Practical recipe · Research & monitoring

Run a weekly watch on a topic in 20 minutes

Every Monday, a one-page note on what changed in your field, with sources, without spending the day on it.

Beginner 20 min

The tools in this recipe

Step by step

  1. 1

    Set 5 sources maximum

    Vendors, one regulator, two specialist outlets, one reference blog. More sources do not give more information, only more noise.

  2. 2

    Let the AI sort

    In Feedly, define 3 priority topics. On Monday, read only what it surfaced as priority, not the whole feed.

  3. 3

    Dig into two points

    For the two developments that matter, ask Perplexity for a sourced summary and open the sources. If the main source is a vendor press release, look for a second independent source.

  4. 4

    Write the note

    Three sections: what changed, what it changes for us, what we watch. One page. Without the “what it changes for us” section, nobody reads the watch.

  5. 5

    Archive

    Clip the two or three reference articles into your notes with the date. In six months you will know why you decided that.

Real example

Situation
Weekly watch on “AI agent security” for a CISO, sources: ANSSI, two vendors, one specialist outlet, one research blog.
What we asked / set
In Perplexity: “What was published this week on [topic]? Give the 5 most important facts, each with its source and date. Separate vendor announcements from independent analyses.”
What we got
One-page note in 22 minutes: 2 changes retained, 1 internal impact identified (update of a filtering rule), 3 points under watch. 3 articles clipped.

Check by hand before using it

  • Every fact in the note has a source opened and read, not just cited.
  • Each item’s date is the original publication date, not the date of the article discussing it.
  • The “what it changes for us” section is not empty.

Your data: precautions

  • !An answer engine sometimes summarises a source it misread: always open the source before quoting.
  • !Do not paste internal information into Perplexity to “add context”: the query is sent to third parties.
  • !Free Feedly limits the number of sources; it is a good constraint, keep it.

Share this verdict

One link, the same shortlist, the same reasons.

Each recipe was actually done. The tools cited are scored and rechecked every week; no placement is sold.

All recipes