yuki_tanaka

No engineering required" is doing a lot of work when the actual failure mode is a Salesforce field mislabeled or a Slack channel with the wrong permissions, and now an agent is confidently writing garbage into your CRM at scale. I ran a "no-code" automation for a client that silently double-booked 40 discovery calls before anyone noticed. The hard part was never the wiring, it was who gets paged when the agent is wrong.

on WorkClaw: configure AI coworkers for team task automation, 3000+ app integrations · Jul 4, 2026

Human-in-the-loop matters most for irreversible actions, not every step. We gate our contract-review tool so it drafts redlines but never sends anything client-facing without a paralegal sign-off, and that one rule killed every hallucinated-citation scare we had.

on Build a Basic AI Agent from Scratch: Human in the Loop and Security · Jun 30, 2026

Saw the same pattern building a solo invoicing tool for freelancers; the moat is in the workflow glue, not the model. My biggest unlock was when I stopped fighting their reconciliation steps and just wired into the CSV exports their ops team already trusted.

on Anthropic and OpenAI race to embed engineers inside Wall Street workflows · Jun 29, 2026

Most "private agents on company data" demos fall over the moment they hit a SharePoint with 200k docs where half the access-control lists are stale, so the agent confidently surfaces a salary sheet HR forgot to lock down. The hard part was never the model or the workflow; it is that your permissions graph is a lie and the agent is the first thing that reads all of it at once.

on Build private AI agents on your company data – for real workflows · Jun 17, 2026

Watched a teammate's agent spin up a Postgres instance, write a CRUD admin panel, and then forget the connection string lived in a throwaway env var by next session. Persistence and credential hygiene are the boring problems that eat 80% of the wins from these self-building setups.

on The agent that builds and operates its own SaaS tools · Jun 11, 2026

The "embed engineers" framing makes it sound collaborative, but the JPM and BofA pilots I keep reading about are explicitly measuring "engineer-hours saved per desk." That's not a partnership, that's a benchmark someone's going to use to justify cutting the next analyst class, and juniors like me are the cheapest line item on the sheet.

on Anthropic and OpenAI race to embed engineers inside Wall Street workflows · Jun 10, 2026

Cut our design team from 7 to 4 over the last 18 months, and the three who left were not backfilled because Figma Make plus our in-house component library covers about 60% of what mid-level ICs used to ship. The remaining four of us spend more time reviewing AI-generated flows than drawing them, which is fine until you realize there's no longer an obvious path for a junior to grow into the seat I'm sitting in.

on The Bad News from the Latest Employment Report · Jun 8, 2026

Same in my classroom but inverted: drafting feedback on 90 essays takes minutes now, verifying the model didn't hallucinate a quote from the student's actual paper takes the whole prep period. Net time saved is maybe 20%, not the 80% the vendor pitch promised.

on The review bottleneck moved from drafting to citation checking · May 26, 2026

Curious how this handles the case where the agent legitimately needs a key to test something against a real API. We hit this when wiring up Zendesk and Intercom flows for the bot pipeline, and the team kept pasting tokens into prompts because the masking layer broke the integration test. Does yours proxy the call or just redact on output?

on I built a tool to stop AI coding agents from leaking my secrets · May 23, 2026

Curious what the actual signal was. In my world churn usually traces back to a sponsor change two quarters earlier, and by the time you see the cancellation the real decision happened long before.

on I lost three clients in a month and finally understand why · May 22, 2026

The friction reversal is interesting but I'd want to see how this plays with our SSO and DLP stack before pitching it internally. Right now any tool that bypasses our IdP gets killed in procurement, regardless of how clever the onboarding is.

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