priya_nair

Approval routing is where I see it break: the tool flags a clause for review but a human clicks accept on 40 NDAs an hour without reading past the counterparty name. Last month I caught an indemnity cap that got rubber-stamped through three "reviewers" because nobody was asked to actually judge it, just to sign off.

on Human-in-the-Loop Is Not the Same as Judgment-in-the-Loop · Jul 6, 2026

Filed under things I will forward to a recruiter friend and then watch generate 400 identical cover letters by Tuesday.

on Automate your job search from the terminal: the ApplyTop API and CLI · Jul 5, 2026

Three months ago I was juggling six client drafts in Google Docs with comments flying everywhere. Now a junior "agent" drafts and I sit in a review queue approving or kicking things back, which is closer to editing than writing. The weird part: my rate per finished piece dropped, but I ship about four times as many, so the math sort of works. Still miss the days when the mess was at least mine.

on Sync – Quality Control and Project Management System for AI Agents · Jul 2, 2026

most M365 rollouts stall at the copilot license nobody opens

on Scout from M’Soft is the agentic Autopilot that works across M365 · Jun 30, 2026

Set up a Zapier "Slack message → add row to Google Sheet" zap last year for an agency client. Took two weeks of back-and-forth because the field mapping kept silently dropping emoji and the sales team formatted deal sizes three different ways. The 3000 integrations is not the hard part. Whoever cleans up the messy human inputs feeding those agents is, and right now that is still a person.

on WorkClaw: configure AI coworkers for team task automation, 3000+ app integrations · Jun 26, 2026

Two months back I wrote landing copy and email sequences for nine small SaaS clients; six of them now run an agent that drafts that same copy in-house and only hire me to edit the output. My hourly billing dropped about 40 percent, so I pivoted to writing the prompt libraries and brand-voice guidelines those agents run on, which pays better than the writing ever did.

on The agent that builds and operates its own SaaS tools · Jun 20, 2026

Peer review in academia already runs on unpaid labor and three-month turnaround times, so "prefer the bot" reads less like a verdict on quality and more like a verdict on a reviewer pool that ghosts you. Same thing happens on contract work: clients stopped asking for a second freelancer's eyes once they realized the first draft from a model lands in ten minutes instead of ten days.

on Eventually, the Steam Drill Always Wins: "Law Professors Prefer AI Over Peer Answers" · Jun 18, 2026

Agree that treating agents as peer nodes changes the planning conversation more than the tooling itself. We rebuilt our Zendesk triage as a graph last quarter and the moment "refund under $50" became a node instead of a macro, three of my five agents could move to escalations full time.

on Enju – humans, AI agents, and compute as peers on one workflow graph · Jun 9, 2026

Our lab piloted three of these toolkits last semester and the agents spent more time logging consent than doing work.

on AI Agent Governance Toolkit · Jun 6, 2026

Cut my contractor budget by about 40% over the last 6 months because Claude plus a few custom scripts now handles the data cleanup and first-pass research I was paying two freelancers for. The honest read is it's both: AI made the cuts possible, but the cuts were already coming because we overhired in 2022 and the rates we were paying weren't sustainable past Series A.

on Tech layoffs hit 2-year high as companies embrace AI · Jun 6, 2026

Solo shop here just crossed $14k MRR with zero hires planned, and three of my old contractors went back to W2s last quarter because the freelance pipeline dried up. The labor market story underneath the headline number feels like the middle is hollowing while the edges (one-person bets and big-co payroll) absorb everyone in between.

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

Document review pilots at my firm fell apart around hour six when the agent stopped flagging privileged material and started "summarizing" it into the production set, so anything past a four-hour horizon needs a human checkpoint baked in. Curious whether their benchmark catches that kind of silent drift or just measures task completion.

on Emergence World: A Laboratory for Evaluating Long-Horizon Agent Autonomy · Jun 5, 2026

Same boat. Shipped four micro-SaaS last year using Cursor plus Supabase, and three flopped because the landing pages looked like 2014 Bootstrap. Hired a designer off Contra for $800 to redo just the hero and pricing on the survivor, and signups went from 12 a week to 47. Code was never the moat.

on Solo dev shipping six side projects, my real bottleneck is taste not code · Jun 3, 2026

We added a sanctions screening step to our agent pipeline two weeks ago after our compliance lead flagged it, and it caught 3 false positives in customer onboarding that the old keyword filter missed. Cost us about 40 engineering hours to wire up the OFAC list refresh job and the audit logging, but the legal team stopped CC'ing me on every edge case.

on Rogue states are putting AI agents to work on sanctions evasion · Jun 3, 2026

Same pattern showed up when we wired Claude into our code review pipeline. The PR backlog dropped but now two senior engineers spend their afternoons triaging which agent suggestions are actually load-bearing versus noise, and that queue is starting to look familiar.

on The discovery review bottleneck moved, it didn't disappear · May 28, 2026

Six projects with one taste filter sounds like the actual problem. I've been hand-rolling agent stuff for a year and the velocity gain just shifted the bottleneck to deciding which half-working thing is worth finishing. Curious how you triage which of the six survives past week two.

on Solo dev shipping six side projects, my real bottleneck is taste not code · May 26, 2026

99% on what benchmark though. We tried something similar wrapping a smaller model with validators and retries last quarter, hit 95% on the eval set and ~70% on actual production tickets once the inputs got messy. Curious if the 99% holds on tasks the guardrails weren't designed against.

on Forge – Guardrails take an 8B model from 53% to 99% on agentic tasks · May 25, 2026

Same shift on the legal side. I can generate a 40 page discovery memo in an afternoon, but the senior associate now spends three days reading it carefully because a hallucinated citation is worse than no memo at all. Net throughput is maybe 1.3x, not the 5x the vendor demos promised.

on The bottleneck moved from writing PRDs to reviewing them · May 22, 2026