2026-07-10
Top stories
- Chat Control 1.0 scanning derogation survives EU Parliament vote European Parliament voted to extend voluntary message scanning on non-encrypted services until 2028 after narrowly failing to reject it.
- Meta ships Muse Spark 1.1 and opens the Meta Model API in public preview Meta released Muse Spark 1.1, a multimodal agentic model with 1 million token context and API pricing that undercuts Grok 4.5.
- Tencent releases Hy3, a 295B open Mixture-of-Experts model Tencent released Hy3, an Apache 2.0 open-weight mixture-of-experts model with 295B parameters and 21B active per token.
- pgrust reimplements PostgreSQL in Rust and passes the regression suite A Rust reimplementation of PostgreSQL passes the full regression suite and boots from existing data directories but is not production-ready.
Conferences and events
- EuroPython 2026 starts in 3 days (2026-07-13) The main annual Python community conference starts 2026-07-13 and covers language, packaging, typing, performance, and ecosystem updates.
- ICML 2026 is active through 2026-07-11 ICML runs through 2026-07-11 with conferences and workshops on training methods, reasoning, evaluation, and interpretability research.
Developer tools
Linux and kernel
Infrastructure
- IERS confirms no leap second at the end of 2026 The International Earth Rotation Service confirmed no leap second will be added at the end of 2026, keeping the UTC offset unchanged.
- Prometheus 3.13.1 fixes a TSDB range-query correctness bug Prometheus 3.13.1 fixes a TSDB bug where the head-chunk cache returned incorrect or missing results on range queries in the LTS line.
Engineering posts
- Good tools are invisible Ginger Bill argues that well-designed tools disappear from focus and prioritize sensible defaults over maximal configurability.
- Write code like a human will maintain it Code quality matters more, not less, when LLMs write it, because poor code trains models to repeat shortcuts in future requests.
Hacker News
- Show HN: colibri streams model weights from SSD to run GLM 5.2 on low-RAM machines A project streams large model weights from SSD to run GLM 5.2 on memory-constrained machines, trading throughput for lower memory usage.
- HN thread on pgrust questions AI-assisted rewrites and test-based reliability HN debated whether databases earn reliability from production exposure rather than passing test suites and questioned AGPL licensing.