“I was never paying for the code. I was paying for the understanding. And some people are about to learn this lesson the hard way.” Josh Wills closes Data Renegades Ep. 13. www.youtube.com/watch?v=gmD1... #DataEngineering #AgenticCoding

“Does Claude need an emotion chip so it can feel shame?” Josh Wills, unprompted, on how engineers build judgment from scars and what the equivalent is for AI agents. www.youtube.com/shorts/6Htkj... #AIAgent #ClaudeCode

“Law school is a pie eating contest where first prize is more pie. And that is true of engineering management as well.” Josh Wills on why he came back to IC work. www.youtube.com/shorts/bTAcw... #EngineeringManagement #DataEngineering

Van Halen buried brown M&Ms in their rider as a canary test -- if the M&Ms were wrong, the pyrotechnics were wrong too. Code used to work the same way. Now it does not. Josh Wills on Ep. 13. www.youtube.com/shorts/-y7pW... #DataEngineering #AgenticCoding

Josh Wills and Claude deleted 120 valuable synthetic datasets while working in tandem. “You let the panic in for about five minutes. Then you get to work.” Data Renegades Ep. 13. www.youtube.com/shorts/q6Ia4... #DataEngineering #AgenticCoding

“You’re here to butter toast. Your job is to run my benchmarks. It is not to interpret things because you don’t have all the context here.” Josh Wills, 18 months into serious agentic workflow use at a frontier lab. www.youtube.com/shorts/IbyjG... #AgenticCoding #AIAgent

Josh Wills coined the 2012 data scientist definition. Now he is IC at DatologyAI building trillion-scale LLM pretraining pipelines. His calibration from 18 months with agents: execution yes, interpretation no. www.youtube.com/watch?v=gmD1... #DataEngineering #AgenticCoding

Netflix experiments took a week to run and failed half the time. Toby Mao migrated them to Spark. An hour. The infrastructure instincts from problems like this became SQLGlot. www.youtube.com/shorts/eG2hy... #DataEngineering #ApacheSpark #DataInfrastructure

“You want a Sketch or a Figma? The market has spoken.” Maxime Beauchemin on collaborative agent platforms vs single-player IDEs. www.youtube.com/shorts/t2zd7... #AgenticCoding #DesignTools #Agor

“Never merge a PR authored by Claude without getting a Codex 5.5 review.” Maxime Beauchemin’s actual multi-model workflow. www.youtube.com/shorts/z8VoI... #AgenticCoding #ClaudeCode #Codex

“There is just about nothing left that you cannot get a better result with an agent.” Maxime Beauchemin, nine months into building Agor. www.youtube.com/shorts/97b86... #AgenticCoding #AIAgent #LLM

"Don't get lazy." Toby Mao on what senior engineers should actually do with AI productivity gains. Hint: more benchmarking, more variation, ruthless review. www.youtube.com/shorts/kYrhH... #AIAgents #SoftwareEngineering #DataEngineering

"nine of the 10 things I tried didn't run faster, but on the 10th attempt, while watching my cooking show, I was able to get significant speed ups." Toby Mao on AI-assisted exploration. www.youtube.com/shorts/0gcfn... #AgenticWorkflows #DataEngineering #AIAssistedDevelopment

“Everyone is a member of technical staff. Your role is to find impact and then automate it and then go find impact again.” Maxime Beauchemin on burning human tokens. www.youtube.com/shorts/WVHx9... #AgenticCoding #FutureOfWork #AIAgent

“I started using cloud code and I was like, my God, this is changing everything.” Maxime Beauchemin on the Claude Code moment and 12 TMUX tabs that became Agor. www.youtube.com/shorts/oq4DV... #AgenticCoding #ClaudeCode #DataRenegades

"My experience at the big companies definitely set me up for a disadvantage." Toby Mao on building for petabyte-scale problems when most companies just use Snowflake. www.youtube.com/shorts/VRP6P... #Startups #DataEngineering #FounderLessons

SQLGlot v30 dropped on the day we recorded. They used Claude to refactor Python into MyPyC-compiled code, got rid of Rust, and hit 5x speed improvements. New episode with Toby Mao. www.youtube.com/watch?v=cREC... #AIAgents #PythonPerformance #DataEngineering

“I cannot see what you’re doing with your agents. I cannot see what I’m doing with my agents.” Maxime Beauchemin names the single-player default of every agentic IDE. www.youtube.com/shorts/rGX1D... #AgenticCoding #MultiAgent #ClaudeCode

"Even the best models today still introduce bugs, one out of every four commits." Toby Mao, creator of SQLGlot, on where the AI line actually is. New episode. www.youtube.com/watch?v=cREC... #AIAgents #LLMCoding #DataEngineering

Maxime Beauchemin created Airflow and Superset. He is back on Data Renegades and the new bet is Agor, a multiplayer agentic-coding platform. www.youtube.com/watch?v=lp97... #AgenticCoding #DataRenegades #Agor

Toby Mao built SQLGlot as a Netflix side project. Now it's inside Google, Airbnb, and Fivetran -- and they just shipped v30 using Claude to refactor thousands of lines into MyPyC-compiled code. 5x faster. No Rust. www.youtube.com/watch?v=cREC...

Ingres was beating Oracle until IBM released DB2 with SQL support. Stonebraker's firsthand account of how one technical choice changed database history. youtube.com/shorts/K7N7Z... #DatabaseHistory #Oracle #Postgres

How IBM Accidentally Made Oracle the Database King -- Mike Stonebraker

How did Oracle become the dominant database? Not because it was better. Mike Stonebraker was there when IBM's DB2 release killed Ingres's momentum and handed Oracle the market. A firsthand account of…

youtube.com

"90% of enterprise programmers are doing maintenance." Stonebraker on where AI should actually be applied. youtube.com/shorts/Crj2M... #AICoding #SoftwareEngineering #LLM

90% of Enterprise Programmers Are Doing Maintenance -- Mike Stonebraker

"90% of enterprise programmers are doing maintenance." Mike Stonebraker says the highest-leverage application of AI isn't writing new code. It's maintaining 30-year-old codebases that have been…

youtube.com

"A trillion dollars is arrayed against us." Stonebraker at 82, betting SQL should orchestrate data queries, not LLMs. youtube.com/shorts/3PPwQ... #AgenticAI #TextToSQL #SQL

A Trillion Dollars Is Arrayed Against Us -- Mike Stonebraker

"A trillion dollars is arrayed against us." At 82, Mike Stonebraker is betting that SQL, not LLMs, should orchestrate enterprise data queries. His research team built Rubicon to prove it. Full…

youtube.com

NoSQL to "not only SQL" to "not yet SQL." Stonebraker's three-word history of a trillion-dollar industry detour. youtube.com/shorts/eATNt... #NoSQL #SQL #DatabaseHistory

The Three-Word History of NoSQL -- Mike Stonebraker

"At the beginning, they said don't use SQL. Then not only SQL. Now? Not yet SQL." Mike Stonebraker's three-word history of the NoSQL movement from the person who called its failure in 2010. Full…

youtube.com

"Don't ever bet against the compiler." Stonebraker on 40 years of database history and why NoSQL converged back to SQL. youtube.com/shorts/eduL-... #SQL #NoSQL #Postgres

Don't Ever Bet Against the Compiler -- Mike Stonebraker

"Don't ever bet against the compiler." Mike Stonebraker won the Turing Award for his work on relational databases. He explains why 40 years of history prove that NoSQL's rejection of SQL was always…

youtube.com

"We tried out the LLMs that everyone else was touting. Accuracy was about 10%, not 80%." Stonebraker on real enterprise text-to-SQL. youtube.com/shorts/iMKre... #TextToSQL #LLM #DataWarehouse

Text-to-SQL Gets 10% on Real Data Warehouses -- Mike Stonebraker

LLMs score 80% on text-to-SQL benchmarks. On MIT's real 1,400-table data warehouse, they score 10%. Turing Award winner Mike Stonebraker breaks down why academic benchmarks hide the real difficulty…

youtube.com

Mike Stonebraker created Postgres, won the Turing Award, and just tested text-to-SQL on real warehouses. The result: 10%, not the 80% benchmarks claim. New episode. www.heavybit.com/library/podc... #TextToSQL #Postgres #DataRenegades

Data Renegades | Ep. #11, Contrarian Bets and AI Skepticism with Michael Stonebraker | Heavybit

On episode 11 of Data Renegades, CL Kao sits down with Michael Stonebraker, legendary database pioneer and creator of Ingres and Postgres.

heavybit.com

Day 1 of #SnowflakeSummit is a lot. Booth visits, back to back talks, so many handshakes. Tomorrow night, relax with us. 🍻 Data Renegades Happy Hour, Tuesday June 2, 6-9pm, 7 min walk from the summit. Drinks, not slides. 😉 luma.com/vnyf1nij?utm...

Data Renegades Snowflake Summit Happy Hour · Luma

In town for Snowflake Summit? Come decompress. Join us for drinks, good conversation, and zero keynote slides. 😉 We're bringing together data practitioners…

luma.com

Data review flagged 99.999% row-count variance. PR was two lines. Base: 5 years of prod history. Current: 1-hour CI build. False alarms train reviewers to scroll past. That's the damage. blog.reccehq.com/session-base... #dbt #DataEngineering

Session Base per PR: Why Data Reviews Lie

Data PR review breaks when the base and current environments are built differently. Here is why, and how session base per PR fixes the false alarms.

blog.reccehq.com