Intelligence Brief
F-Droid 2.0 shipped 2026-09-25 after over a year of development and 14 test releases
F-Droid 2.0 shipped 2026-09-25 after over a year · F-Droid 2.0 also removed two security features rather · Browser-use's own co-founders
1067
points on "F-Droid 2.0"
Today's Insights
F-Droid 2.0 shipped 2026-09-25 after over a year of development and 14 test releases — a full Kotlin Compose rewrite, its biggest change in a decade — landing six days before Google's 30 September deadline extending mandatory developer-identity verification to sideloaded APKs on certified devices.
F-Droid 2.0 also removed two security features rather than adding them: Tor auto-detection (replaced by a Tor-VPN recommendation) and the app-wiping panic response — a maintainability trade-off, not a hardening move, the same release cycle as the Play verification deadline.
Browser-use's own co-founders, Gregor Žunič and Magnus Müller, were both actively engineering the org's benchmark repo today — three weeks after BrowserCode v2 reported GPT-6 Astra at 77.3% versus Claude Opus 5 at 50.5% on browser-agent tasks, a gap still unreplicated by any independent lab.
Georgi Gerganov's llama.cpp shipped three tagged releases in 30 minutes today, one adding q5_k quantization for Qualcomm's Hexagon NPU backend — extending on-device inference's quantized-release cadence past CPU/GPU into dedicated mobile silicon.
Google confirmed a Project Suncatcher prototype satellite carrying its own TPUs launches October 1 on a SpaceX Falcon 9 — the first in-orbit test of whether AI chips survive space radiation, vibration, and vacuum cooling for future space-based compute.
Simon Willison shipped commit-rewriter 0.2 (non-default-branch support) and datasette 1.0a41 (OpenTelemetry, unified modal Web Component) within about 90 minutes of each other on 2026-09-24 evening — a second dated instance of the same-day multi-ship cadence this vault flagged 2026-09-14 as a falsification test.
Today's Google Play top-30 (Portugal-localized capture) shows no ChatGPT, Claude, or Perplexity — continuing the AI-chat-app absence pattern this vault has treated as scrape-window noise since early August rather than a genuine demand signal.
Original agent-framework authors remain active maintainers rather than having moved on: Chi Wang, Qingyun Wu (AutoGen), Toran Bruce Richards (AutoGPT), and Dify's maintainers all pushed or closed issues on their own frameworks in the last 24 hours — a durability signal for a category this vault has tracked since January.
Trending Repos
- vectorize-io/hindsight
Python
+1,668/d - +1,373/d
- dream-num/univer
TypeScript
+1,082/d - obra/superpowers
Shell
+611/d - +509/d