What the people actually building AI said today. One page — a 1-min read.
Arca showed what it takes to run AI agents through financial work that lasts days: persistent client context, checks on every action, and human approval for trades. Aaron Levie argued that AI should raise the quality of the work itself, while Swyx made a case for faster critique loops over maximum model effort.
Arca’s agents coordinate financial work across days, with humans approving trades
Arca uses agents to analyze client finances, track work across days, and bring in an advisor when a decision needs human judgment. “Agents prompt people more than people prompt agents.” The firm keeps trades and client communications under human control, while logging actions and checking agent output for errors and compliance issues.
Aaron Levie says AI work should exceed what teams could do before
Levie’s test for effective AI use is whether the result would have been possible before AI. He argues that teams should aim for better product features, faster engineering systems, deeper research, and business processes that learn from past failures, rather than simply finishing familiar work faster.
Swyx favors low model effort paired with a critique loop
For work limited by throughput, Swyx argues that low effort model runs followed by critique can beat pushing every run to maximum effort. He treats effort settings as a matter of taste rather than a universal rule.
Source data comes from the open-source project follow-builders by zarazhangrui, released under the MIT license. Summaries are generated by an LLM from that project's public feeds, and the summarization prompts are adapted from it. Every item above links to its original source.
Summaries generated automatically. Read the original before relying on any claim.
