AI coding only feels identity-threatening when the goals stay too small
doodlestein · x · 2026-07-23
The author argues that AI makes coding and math feel less identity-defining only if you keep doing low-difficulty work.
- His answer is to raise ambition sharply: run 40–50 projects in parallel and aim for problems that once required PhD-level teams and years of effort.
- He says his newer projects are now built on the foundations of earlier ones, creating a compounding effect.
- For this kind of scale, he recommends a different workflow, and says the only one he’s seen work is the Agent Flywheel.
Related event: Opinion: Counter AI Impact with Greater Ambition and Agent Workflows(2 posts)→
More from coding & agent
- Five frontier LLMs hallucinated the same fake package names in 200,000 coding prompts — nptacek · 2026-07-23
- Matt Shumer Shares Multi-Agent Workflow: Running an AI Company via Workbench — mattshumer_ · 2026-07-23
- Dev Embeds OpenAI Codex in the Browser via WebAssembly — cocktailpeanut · 2026-07-23
- Open Source Project Unifies All RL Environments Behind a Single API — eliebakouch · 2026-07-23
- Matt Shumer is running a company of agents across laptops, cloud VMs and one chief of staff — josh_bickett · 2026-07-23
- SubText to debut as a new file format and LM tooling stack for text — HamelHusain · 2026-07-23