Support The Automated Daily directly:
15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going:
theautomateddaily.com/support
Please support this podcast by checking out our sponsors:
- Consensus: AI for Research. Get a free month -
https://get.consensus.app/automated_daily
- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time -
https://get.surveymonkey.com/tad
- Effortless AI design for presentations, websites, and more with Gamma -
https://try.gamma.app/tad
Today's topics:
AI code review delays PostgreSQL 19 - PostgreSQL 19 has been pushed to around November 19 after AI-assisted code review found so many defects that developers reverted about eight to ten major features, including SQL Property Graph Queries. Those features are expected to move to PostgreSQL 20.
Meta's Muse Spark Open runs locally - Meta released Muse Spark Open, an 816-billion-parameter open-weight multimodal model with a 1 million-token context window. Meta says it can run locally on two Apple M5 Ultra Mac Studios using MLX.
NVIDIA Nemotron 3.5 Super VL - NVIDIA introduced Nemotron 3.5 Super VL, an open mixture-of-experts multimodal model built for enterprise AI agents. It handles text, images, documents and video, and comes with NeMo datasets and tools for customization.
FLUX 3 Image Klein open weights - Black Forest Labs released FLUX 3 Image Klein, a 5.6B open-weights text-to-image and image editing model. It supports native 4K output and multi-reference generation, and is designed to run on consumer GPUs.
Vercel index: open models surge - Vercel's AI Gateway Production Index reports gateway spend up about 13x since December while the price per token fell 47%. Open-weight models now handle more than 71% of tokens, and teams switch AI providers often.
How OpenAI's Codex turned around - A look back at how OpenAI's Codex went from a weak cloud coding tool to a key product. The turnaround came from reasoning models, autonomous agentic workflows and a move to the terminal, prompted in part by competition from Claude Code.
Perplexity open-sources SPACE agent sandbox - Perplexity is open-sourcing SPACE, the sandbox platform behind its Computer product. It provides isolation, state recovery, and controls over credentials and network access for autonomous AI agents.
Anthropic embeds METR evaluators - Anthropic is giving a METR team employee-like access to independently evaluate its models and safety practices, and METR will publish its risk assessments. SecureBio, Accenture and Faculty are working with Anthropic at lower access tiers.
Benchmarks miss most real work - Mercor's Benchmark Map finds that AI benchmarks cover only 208 of 923 O*NET occupations. About 90% of US working hours go to tasks with little or no benchmark coverage.
Policy gradients as covariance - A research post shows that the expected policy-gradient update in reinforcement learning is a covariance between reward and score, with Fisher information as the matching variance. It also gives a centered
