Track 1 vs Track 2: the data science work AI is commoditizing, and what stays valuable
mdancho84 · x · 2026-09-22
Data science influencer mdancho84 argues in a thread that 'Track 1' work — running analyses, reading docs, summarizing findings, testing ideas manually — is being rapidly commoditized by AI, while 'Track 2' people who architect the systems (agents, workflows, memory, tools, evaluation loops, business integration) stay valuable.
He cites Karpathy's framing: frontier AI research used to be done by 'meat computers' meeting via 'sound wave interconnect'; it now belongs to autonomous swarms of AI agents. Reading papers, generating ideas, writing experiments, evaluating, iterating — the process once called research is turning into software. AutoResearch isn't just another cool repo, it's a signal.
Related event: AI Threatens to Commoditize Data Science Work(2 posts)→
More from AGI Musings
- Dartmouth Provost's op-eds flagged as 96% AI-written by near-zero-error detector Pangram — soumitrashukla9 · 2026-09-22
- Economists: falling labor share and rising real wages aren't in tension in the AI era — soumitrashukla9 · 2026-09-22
- Chatbots Were the Warm-Up: Why World Models May Underpin Physical AI — ingliguori · 2026-09-22
- Creator's rule for AI: delegate execution and retrieval, never surrender the thinking — romitheguru · 2026-09-22
- 39 peer-reviewed papers on catastrophic risk from AGI, systematically listed — ben_j_todd · 2026-09-22
- COO of multi-billion-dollar manufacturer: all big-data LLM pilots cancelled — richdotca · 2026-09-22