tim’s radarVerified news /
November 12 digest: decision APIs, Tabletop and Rust safety
Three checked official sources, access limits and ten historical GitHub observations. No independent product or security tests.
Three things to read first
Today’s selections
A dedicated decision endpoint
Google introduced an API that scores bounded choices across multimodal inputs and returns confidence values. It supports yes/no, multiple-choice and rubric questions.
Evaluate it on labeled examples before using confidence values to automate routing; keep a human fallback. We did not test its accuracy.
Source and verification limits
Checked: 2026-11-12
Announcement checked; calibration and performance not independently tested.
Tabletop: check eligibility
Google Labs announced Tabletop, an experimental shared space that generates interactive planning tools from links, files and prompts.
Check eligibility first: the rollout is through a U.S. waitlist for adults aged 18 and over. Do not assume Taiwan access. We have not tried it.
Source and verification limits
Checked: 2026-11-12
U.S. adults via waitlist; no independent product trial.
Security watch
Rust safety still needs review
Microsoft reports AI-assisted investigation of soundness issues in Rust standard-library unsafe code beneath safe APIs. Researchers reproduced and reported the findings; some issues remained open at publication.
Review the linked upstream issues for relevance to your code. This research is not evidence that every Rust application is exploitable; we did not reproduce the experiments.
Source and verification limits
Checked: 2026-11-12
Research report checked; no reproduced experiments or universal patch claim.
GitHub momentum
Momentum is not a recommendation. Descriptions are project claims, not independently tested. Counts are historical snapshots, not live values.
Agent skill collection
- Total stars
- 5,895
- Since previous daily snapshot
- 749
- 7-day gain
- —
- 7-day commits
- 0
Customizable design system
- Total stars
- 25,039
- Since previous daily snapshot
- 611
- 7-day gain
- 11,199
- 7-day commits
- 1,074
Virtual-machine environment for Claude
- Total stars
- 3,298
- Since previous daily snapshot
- 556
- 7-day gain
- —
- 7-day commits
- 239
Native game recompilation project
- Total stars
- 7,839
- Since previous daily snapshot
- 549
- 7-day gain
- —
- 7-day commits
- 1,970
Software analysis and recreation engine
- Total stars
- 28,462
- Since previous daily snapshot
- 461
- 7-day gain
- 16,771
- 7-day commits
- 572
Download manager
- Total stars
- 6,860
- Since previous daily snapshot
- 361
- 7-day gain
- —
- 7-day commits
- 292
iOS MobileGestalt editor
- Total stars
- 6,270
- Since previous daily snapshot
- 265
- 7-day gain
- —
- 7-day commits
- 130
Computer-use command-line tool
- Total stars
- 22,104
- Since previous daily snapshot
- 237
- 7-day gain
- 7,054
- 7-day commits
- 9
Agent-assisted software design plugin
- Total stars
- 17,699
- Since previous daily snapshot
- 233
- 7-day gain
- —
- 7-day commits
- 45
Interface design refinement skill
- Total stars
- 14,563
- Since previous daily snapshot
- 230
- 7-day gain
- 5,900
- 7-day commits
- 16
Scope and verification
Selected from November 12 snapshots at 09:29–09:31 Taiwan time. Publication dates follow original pages, not RSS labels. Verified means sources support these summaries, not independent testing.
Descriptions are project self-descriptions, not tests. Daily gains compare stored daily snapshots about 10 hours apart, not a precise 24 hours. Google DeepMind RSS failed; selected official pages were read directly. Images omitted because asset-specific permission and authorship were not verified this run.
Sources
- Gemini Decisions API for bounded choices · 2026-11-12
- Tabletop starts with a limited rollout · 2026-11-12
- FORGE investigates Rust library soundness · 2026-11-12