g_r_e_g_r_e_g_r's profile picture. Co-founder @ZerveAI | Building the future of agentic data science | AI, infra, & workflow automation | Human in the loop, by design

Greg Michaelson

@g_r_e_g_r_e_g_r

Co-founder @ZerveAI | Building the future of agentic data science | AI, infra, & workflow automation | Human in the loop, by design

LLMs are commodities. What separates coding agents now is everything around the model: project memory, task-aware retrieval, runtime loop (write→run→observe→fix), tool/infrastructure integration, and transparent UX. In data science, workflow > model. #DataScience #AI


Check out the latest random assignment of buzzwords to arbitrarily curvy lines. Thanks to pay-to-play B2B pirates for the informative post! gartner.com/en/newsroom/pr…


LLMs aren't ready to one-shot data science—yet. Biggest limitation: They can't see your data. Quick tips: -Share data snapshots -Request hypotheses -Get targeted analysis tips -Regularly re-summarize progress to "extend" context Future: Agentic AI that explores data directly.


LLMs have a Dunning-Kruger problem: they confidently answer every question, regardless of accuracy—but quickly lose confidence when challenged. To improve LLMs, we must calibrate their certainty realistically, not just increase accuracy. #LLMs #AI #DunningKruger #GenerativeAI

g_r_e_g_r_e_g_r's tweet image. LLMs have a Dunning-Kruger problem: they confidently answer every question, regardless of accuracy—but quickly lose confidence when challenged. To improve LLMs, we must calibrate their certainty realistically, not just increase accuracy. #LLMs #AI #DunningKruger #GenerativeAI

United States Trends

Loading...

Something went wrong.


Something went wrong.