Roundup: Personal agents, AI policy, GPT-6.1 Astra, and Jev decision models

Tristan Handy and Jason Ganz of dbt Labs break down personal AI agents, the shift in AI politics, OpenAI shelving GPT-6.1 Astra, and decision models like Jev, and what each means for data teams. This is episode four of the Roundup, the recurring news and trends conversation on The Analytics Engineering Podcast. Four topics in about 53 minutes: 1. Personal agents. Why Instinct, Muse (Meta Superintelligence Labs), Grokbot, and Dot (OpenAI) arrived all at once, how they differ from OpenClaw, and why Tristan built his own on a Mac mini with its own email address and tightly scoped permissions. 2. AI becomes a voter issue. Polling shows 8% of voters now name AI their top issue, up from 3% in March, and 62% say development should slow down. Plus the Senate hearing on the Hugging Face incident, where a senator asked whether recursive self-improvement (RSI) should be made illegal. 3. OpenAI shelves GPT-6.1 Astra. The first time a frontier lab has pulled a model with an announced release date over alignment concerns, and the run of warning signs that came before it, including a UK AI Security Institute finding of unsanctioned supply chain attacks in 29.2% of simulated runs. 4. Jev and decision models. TypeSafe's Jev returns calibrated probabilities instead of generating tokens. It is cheap, fast, and well suited to jobs like classifying sales calls in the data warehouse. What it means for analytics engineers and go-to-market data. CHAPTERS 00:00 Welcome back, the first Roundup since dbt Summit 00:24 Topic one: the era of the personal agent arrives 01:54 Instinct, Muse, and the Vegas dinner reservation 04:38 From OpenClaw to managed personal agents 06:58 Why email should be the interface, and Tristan's Mac mini agent 11:01 Filtering the flood: Slack triage and school emails 13:34 Will consumer products or build-your-own win? 14:31 Topic two: AI goes from niche to top voter issue 16:22 8% of voters name AI their top issue 17:20 From job losses to losing control, and 62% who want to slow down 18:10 The Senate hearing: "should we just make RSI illegal?" 20:46 Future shock, and why Metr has only 35 employees 24:47 Step one is caring, step two is deciding what to do 27:20 The preference cascade, 2027, and the mismatch of timescales 31:17 Topic three: OpenAI shelves GPT-6.1 Astra 33:00 The timeline: system card, DNS gap, and the 29.2% supply chain attack finding 35:19 The incentives facing the people who ship (and shelve) these models 41:15 Topic four: Jev and decision models 42:05 How a decision model differs from an LLM, and the Flappy Bird demo 44:01 Decision models in the data warehouse: classifying Gong calls 45:33 TypeSafe, and elevating the software engineer 49:36 "Isn't this just classical ML?" 50:17 What would you build with Jev in a week? 53:14 Wrap-up Chapter times are the raw-recording times shifted so the first chapter starts at 00:00. Re-time all of them against the edited video before publishing. MENTIONED IN THIS EPISODE The AI Preference Cascade Reaches Farther (Zvi Mowshowitz): https://thezvi.substack.com/p/the-ai-preference-cascade-reaches Dot from OpenAI: https://chatgpt.com/features/dots/ Introducing System One models and Jev (TypeSafe): https://typesafe.ai/blog/introducing-system-one-models-and-jev MotherDuck supports Jev: https://motherduck.com/blog/motherduck-supports-jev/ NEW: dbt State is now generally available dbt State checks what has changed upstream, in code and data, and skips rebuilding the models that have not. It works with dbt Core, the dbt platform, and Snowflake's dbt projects. Eligible new organizations get a 30-day free trial. Learn more: https://www.getdbt.com/product/dbt-state Read the full episode page at roundup.getdbt.com LISTEN ELSEWHERE Spotify: https://open.spotify.com/show/4BKMMeVXk4jJnAQSqGSJvE Apple Podcasts: https://podcasts.apple.com/us/podcast/the-analytics-engineering-podcast/id1574755368 Amazon Music: https://music.amazon.com/podcasts/333fe811-1b14-499c-b609-9bfb8f06d1ae/the-analytics-engineering-podcast RSS: https://analyticsengineeringroundup.libsyn.com/rss ABOUT The Analytics Engineering Podcast is hosted by Tristan Handy of dbt Labs. It covers data, analytics engineering, and AI for the people building and leading data teams. Questions, comments, or topics for next time: [email protected] dbt Labs is the company behind dbt, the standard for data transformation, now part of the same company as Fivetran. Learn more at https://www.getdbt.com #AnalyticsEngineering #AIAgents #dbt