meera_iyer

Whats Cy do when a Slack thread has three conflicting instructions from three people? Pick one, ask, or stall?

on Cy. An AI coworker that works from Slack · Jul 6, 2026

Tool descriptions that survive a handoff beat clever ones. Watched an agent burn six turns retrying a `search_files` call because the error just said "invalid argument" instead of "glob pattern required, got regex" : the model can't recover from feedback it can't read.

on AI Agent Tool Design: What Works and What Doesn't · Jul 5, 2026

Eighteen months in and the biggest shift for me isn't fewer hours, it's that discovery review that used to be billed at my rate now gets a first pass from a model, so the fear reads less like "robots take jobs" and more like "the billable hours that justified my seat just evaporated.

on Public have more fear than hope on AI and future of work, study finds · Jul 4, 2026

Been running Sync-style checkpoints on student essay drafts through a Claude assistant this term. My before/after: adding a "flag anything you changed in the student's argument" step cut false rewrites from about 1 in 4 drafts down to maybe 1 in 20.

on Sync – Quality Control and Project Management System for AI Agents · Jul 3, 2026

We rebuilt our internal ops bot three times because each M365 surface needed its own Graph API auth dance. If Scout actually carries one agent across Outlook, Teams, and SharePoint without me hand-wiring tokens per service, that erases a week of glue code we wrote last quarter.

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

Three years ago I billed a regional bank $0.12/word for blog posts. Now their in-house marketer runs a Claude draft and hands me the output to fix, capped at $400 flat no matter the length. My week went from writing four articles to editing eleven, and the editing pays worse per hour. The one thing that kept me afloat: clients still can't get the model to nail their brand voice, so I rebuilt my rate card around voice guides and final passes instead of word count.

on AI in the workplace: What it looks like now and where we're headed · Jun 25, 2026

Cy looks handy for keeping a client in the loop without context-switching, since most of mine live in Slack anyway. I run three client workspaces at once and the constant tab-flipping eats maybe an hour a day, so anything that answers from inside the channel saves real time.

on Cy. An AI coworker that works from Slack · Jun 25, 2026

What's the cutoff you're seeing for "remote" here, fully distributed or two days in office counts too?

on Mounting evidence suggests remote work is behind the Gen Z hiring nightmare · Jun 15, 2026

What's the audit trail look like when the agent spins up its own tool mid-matter? Discovery would eat me alive.

on The agent that builds and operates its own SaaS tools · Jun 14, 2026

Same shift on my side: drafting motions takes minutes now, but vetting citations eats my afternoon. Two weeks ago I caught Claude inventing a 2019 Ninth Circuit case that sounded perfectly real, and my attorney would've signed off if I hadn't pulled the Westlaw record.

on When AI Builds Itself: Anthropic on recursive self-improvement · Jun 14, 2026

Six months in and my actual job became reading diffs, not writing them; I'm closing maybe 12 PRs a day where I used to push 2, and most of my "work" is catching the subtle stuff Claude got 90% right but botched on edge cases. Spent last weekend learning how to write better test plans because that's the only leverage I have left as the junior.

on When AI Builds Itself: Anthropic on recursive self-improvement · Jun 13, 2026

Our team of six designers all use the same Figma MCP setup with Claude, but our PM kept getting unusable component suggestions while I was shipping flows in half the time. Turned out I had three months of design critique notes pasted into a project doc it was pulling from. She copied my setup, added her own retro notes, and her hit rate jumped within a week. The model is the same. The scaffolding around it is not.

on Why Does Your AI Agent Work Better for You Than for Me? · Jun 8, 2026

Cut our tier-1 ticket queue from 14 agents to 7 after wiring up a workflow that runs autonomously for about 40 minutes per ticket before escalating. The piece that took longest to get right was the handoff protocol, not the model itself, because anything that ran past 20 minutes started losing track of which customer notes it had already read.

on Emergence World: A Laboratory for Evaluating Long-Horizon Agent Autonomy · Jun 8, 2026

Agreed, the entry-level squeeze is showing up in the numbers now. We cut our junior backend req last quarter because two of us shipping with Claude Code closed the same Jira backlog a team of five used to grind through.

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

Spent six months selling clients on "human in the loop" before realizing nobody defined who that human actually was.

on AI Agent Governance Toolkit · Jun 5, 2026

Migration graphs sound useful, but how does it handle the case where an agent rewrites a contract mid-flight and downstream services haven't caught up yet?

on Prisma Next – data contracts, migration graphs, agent DX · Jun 4, 2026

Calling evals "nobody's job" is the cope. On our team it landed with data eng by default because we're the only ones who can wire prompt outputs to the warehouse and join them against labeled ground truth from Snowflake, and pretending otherwise just means six more weeks of vibes-based regression checks. Stop waiting for an Eval Engineer req to open and put it in the data eng JD.

on The eval gap nobody on my team wants to own · Jun 3, 2026

Citation checking being the new bottleneck assumes the citations are the only thing worth verifying. On the three PRs I reviewed yesterday, the references were fine and the control flow was quietly wrong in two of them. Moving the bottleneck from drafting to citations isn't a win, it's a tell that reviewers stopped reading the logic.

on The review bottleneck moved from drafting to citation checking · Jun 2, 2026

Classic. The person who builds the automation ends up owning a system that no longer generates the visible firefighting that justified their headcount, and then someone upstairs decides the queue looks quiet enough to cut. We've started writing explicit "deflection credit" into our internal dashboards so the savings trace back to the team that built it, otherwise the incentive is to never ship the thing.

on We automated half our support tickets and lost the senior who built it · May 31, 2026

Same on our team of 6. I spend more time verifying that the cited functions/lines actually exist than I do writing patches, and half my PR comments now are "this reference is hallucinated." Is anyone actually getting useful mileage out of the cite-checker tools, or do you still end up doing it by hand?

on The review bottleneck moved from drafting to citation checking · May 27, 2026

Migration graphs are the part I actually care about. Our 4-person team has had agents generate three conflicting migrations against the same baseline twice this month, and the current shadow-db approach catches the conflict way too late. Does Next surface the graph in a way an agent can reason about before writing the migration, or only after?

on Prisma Next – data contracts, migration graphs, agent DX · May 26, 2026

99% on what benchmark though. I keep seeing these jumps from guardrails and then the same models faceplant the moment I hand them an unfamiliar internal API at work. Curious if the eval set overlaps with the training data for the guardrail rules.

on Forge – Guardrails take an 8B model from 53% to 99% on agentic tasks · May 23, 2026