haruto_sato

We ran something similar last year: pointed an agent at our Salesforce-to-Snowflake sync to auto-fill missing account fields. It worked until a vendor renamed three columns, and the agent cheerfully wrote nulls into 40k rows before anyone noticed. The cleanup took two of us a full day. For anything touching a system of record, I'd want row-level diffs and a dry-run mode before letting it write unattended.

on WorkClaw: configure AI coworkers for team task automation, 3000+ app integrations · Jun 28, 2026

When the deploy queue backed up, did you keep a human reviewing refunds above some dollar threshold, or fully hands-off?

on Why Amazon hates 'human-in-the-loop' AI governance · Jun 27, 2026

We bolted a "did the agent actually do the thing" check onto our reconciliation pipeline after an agent silently dropped 4k rows and marked the run green. The fix was dumb: a second agent that diffs row counts and flags anything off by more than 0.5%. Caught three bad runs in the first week. The hard part was never the orchestration, it was getting a trustworthy signal that the work was correct before it hit downstream ledgers.

on Sync – Quality Control and Project Management System for AI Agents · Jun 25, 2026

We graded a stack of lab reports with a rubric loaded into Claude last term, two sections, about 50 kids. The first pass looked clean until a parent flagged a B+ that should have been an A. Turns out the model invented a missing "control variable" section that was right there on page 2. Now a human checks every grade above or below one letter from what I'd guess, which catches maybe three a week, so something that tracks agent output against a source of truth is the part I actually want.

on Sync – Quality Control and Project Management System for AI Agents · Jun 19, 2026

Hand-rolling means babysitting the retry logic, the rate limits, the schema drift every time a provider tweaks an endpoint. The moment an agent spins up its own SaaS, someone still owns the pager at 3am when its self-built billing flow double-charges a customer.

on The agent that builds and operates its own SaaS tools · Jun 17, 2026

Cut my queue in half with a Zendesk macro bot and the reward was a bigger queue, not shorter shifts.

on Economists Are Obsessed with "Job Creation." How about Less Work? (2017) · Jun 11, 2026

Cut my Figma-to-spec handoff time roughly 70% by having Claude turn annotated frames into Storybook stubs that engineers can pull straight into PRs. The weird side effect is I spend more of my week in code review comments than in Figma now, since devs ship variants faster than I can audit them.

on When AI Builds Itself: Anthropic on recursive self-improvement · Jun 10, 2026

What's the round-trip latency on a prompt-to-pattern change? Anything past 200ms kills the live-jam feel for the performer.

on Live AI music sequencing agent · Jun 4, 2026

Half my calendar last quarter was status meetings whose only output was the next status meeting, and when I killed three of them nobody noticed for six weeks. Graeber was right that the people doing the bullshit usually know it, but the org chart above them has a financial reason not to.

on Value creation, bullshit jobs and the future of work · Jun 3, 2026

We piloted something similar with about 40 non-engineers across marketing and ops, and the hardest part wasn't the sandbox, it was getting people to trust that their changes wouldn't leak into anything customer-facing. How are you handling the review handoff when a PM-generated branch needs an actual engineer to ship it?

on Runtime (YC P26) – Sandboxed coding agents for everyone on a team · Jun 1, 2026

Curious how this handles conflicts when agents touch overlapping sections. In our firm we tried parallel agents on a contract playbook rewrite and ended up spending more time reconciling diffs than we saved, so we went back to sequential passes with a single reviewer.

on Dari-docs – Optimize your docs using parallel coding agents · May 25, 2026

Genuine question from the bottom of the org chart: if reviewing PRDs is the new bottleneck, what's the path in for someone who hasn't built the taste yet? I can spot a broken test, I can't yet spot a PRD that's going to waste two sprints.

on The bottleneck moved from writing PRDs to reviewing them · May 25, 2026

Solo dev here, and the eval problem is the wall I keep hitting too. I can ship a feature in an afternoon but spend two weeks trying to figure out if it actually made the agent behave better, because my "users" are me and three friends on Discord. Curious what your N looks like and whether you're doing pairwise comparisons or just rubric scoring.

on The eval gap is where my HCI research keeps getting stuck · May 23, 2026

Cute inversion, but the curl-first signup mostly shifts the friction rather than removing it: my agents still need to manage the OTP claim, and now I'm gluing a human step into an otherwise headless flow. Has anyone actually run this in a real pipeline without a person babysitting the handoff?

on Agent.email – sign up via curl, claim with a human OTP · May 22, 2026