Pondero Brief - OCTOBER 20TH, 2026
Defender now flags 60 agent types, per Microsoft. Plus Nvidia's 74B agent model in 24GB.
|
|
|
Guardian Agents vets every agent action before it runs, and risky ones wait for a human sign-off.
|
|
Microsoft's security chief says frontier models "cannot be trusted"; Defender just added 35 agent types, 60 total including Claude Code, per Microsoft. Read why Microsoft wants agent monitoring kept apart from the model vendor.
|
|
|
|
|
|
This week we started something new. Pondero Labs puts AI tools through real jobs and publishes everything: the setup, the screenshots, the answer key and the raw results.
No vendor demos and no paid results. We write the answer key before we run a test, so you can check our grading or rerun it yourself.
The first five labs:
See every test on the Pondero Labs page. Want a tool tested next? Reply to this email and tell us which one.
|
|
|
|
|
Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.
|
|
|
Affiliate disclosure
·
Unsubscribe
·
Manage preferences
Pondero earns commissions on some links. This does not affect our editorial picks.
|
|
