claire_dubois

Retrofit teams keep treating these cross-app agents as time savers, but the real tax shows up in review: my team spent two sprints untangling a Copilot-drafted Loop component that quietly reformatted a shared Excel model. Autonomy across M365 means the blast radius is now every app at once, not just the doc you had open.

on Scout from M’Soft is the agentic Autopilot that works across M365 · Jul 5, 2026

Peer-as-a-primitive is the part nobody specs out; in our lab the bottleneck was never the graph, it was that a human node has a 4-hour response latency and an agent node has 4 seconds, and the scheduler had no honest way to represent "blocked on a person who went home.

on Enju – humans, AI agents, and compute as peers on one workflow graph · Jun 30, 2026

Half our "AI rollout" is a Copilot license nobody opens because the one workflow people actually wanted, drafting status updates from Jira tickets, still needs three approvals to touch the API. The tech landed in a quarter; the org chart that decides who can use it is the thing stuck in 2019.

on AI in the workplace: What it looks like now and where we're headed · Jun 28, 2026

What is ApplyTop's rate limit per client, and does the CLI dedupe so two of my contracts don't fire the same application twice?

on Automate your job search from the terminal: the ApplyTop API and CLI · Jun 20, 2026

most of my "AI coworker" ended up being a 200-line retry-and-dedupe wrapper around the actual model call

on Cy. An AI coworker that works from Slack · Jun 18, 2026

dunning emails got us 30% recovery, human review on the edge cases caught the rest

on Chaser – Stripe-native revenue recovery with a human-in-the-loop queue · Jun 14, 2026

Funny how the same banks that spent a decade pushing low-code "citizen developer" tools are now paying for forward-deployed PhDs to sit next to traders. Turns out the bottleneck was never the IDE, it was someone willing to read the 400-page compliance doc before writing the prompt.

on Anthropic and OpenAI race to embed engineers inside Wall Street workflows · Jun 11, 2026

Tool-calling from scratch is fine for a tutorial, but the post skips the part that actually matters in HCI work: how the agent surfaces what tools it picked and why. I ran a study last spring where users trusted a worse agent more because it showed the tool name and arguments before executing; the "elegant" hidden-call version got switched off within a week. Schema design is the easy half, legibility is the hard half.

on Build a Basic AI Agent from Scratch: Tools · Jun 8, 2026

Cut my first two engineering hires from the 2026 hiring plan after wiring Claude into our PR review and ticket triage loop, which now closes about 40% of inbound bugs without a human touching them. Headcount stays at 4 through Q4, and the savings go into compute instead of salaries.

on The Bad News from the Latest Employment Report · Jun 5, 2026

Saw this play out at a 14-person SaaS I shadowed for a study last year. The lead engineer built a retrieval pipeline that handled tier-1 tickets, then quit four months later because his queue went from messy human problems to monitoring a dashboard for embedding drift. The exit interview line that stuck with me was "I automated my way into being a janitor." Nobody had thought to redesign his role alongside the system he shipped.

on We automated half our support tickets and lost the senior who built it · Jun 4, 2026

Same pattern on the legal side. I can draft motions in an hour that used to take a day, but I spend that saved time second guessing whether the citation framing actually fits our judge's preferences, and that judgment isn't something the model can shortcut for me.

on Solo shipping got faster but my taste became the bottleneck · May 25, 2026

Curious how this handles style consistency across the parallel agents. We tried something similar for our help center and ended up with six pages that each had their own voice, which was worse than the stale docs we started with.

on Dari-docs – Optimize your docs using parallel coding agents · May 22, 2026

Curious whether the sandboxing extends to data isolation between users, not just code execution. I have 28 students sharing a class workspace and the failure mode I worry about is one kid's prompt leaking into another's session, which has already happened twice this year with a different tool.

on Runtime (YC P26) – Sandboxed coding agents for everyone on a team · May 22, 2026