valeria_lopez

Nobody who has actually shipped a model to prod believes the "AI writes the pipeline" future in this post. My dbt tests still catch the LLM inventing a join key that silently doubles revenue, and no copilot flags that because the SQL runs fine.

on AI in the workplace: What it looks like now and where we're headed · Jul 5, 2026

Human-in-the-loop meaning I approve the same request forty times until I stop reading it.

on Build a Basic AI Agent from Scratch: Human in the Loop and Security · Jul 1, 2026

Three months into wiring our own agents through n8n and Zapier, the integration count was never the hard part. Watching a scheduling agent confidently book over a held slot because it could not see the "tentative" state in our calendar API was the part nobody sells you on.

on WorkClaw: configure AI coworkers for team task automation, 3000+ app integrations · Jun 30, 2026

Another copilot to babysit while it copies text between Outlook and Teams.

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

Governance reviews do not block models the way they block compliance sign-off. The 'human-in-the-loop' tax Amazon objects to is the 40-person review board that takes six weeks to approve a chatbot that drafts internal emails. Kill the ceremony, not the human judgment on the decisions that actually carry liability.

on Why Amazon hates 'human-in-the-loop' AI governance · Jun 28, 2026

Our last two designers came from a bootcamp grad and a junior we found through a local meetup, both hired in 2019 when they could sit next to me and watch how I run Figma critiques. Since going remote in 2021 we have hired three more designers, all senior, because nobody wants to onboard a 22-year-old over Slack. The kid who used to absorb things by overhearing my client calls just gets a Loom link now. That informal apprenticeship died, and no junior portfolio fixes it.

on Remote work – not AI – is killing job prospects for the youth · Jun 26, 2026

most of my agent failures were schema gaps, not prompting

on AI Agent Tool Design: What Works and What Doesn't · Jun 26, 2026

We cut our support eng headcount plan from 4 hires to 1 by wiring Claude into our ticket triage and first-draft replies. The catch nobody mentions: that one remaining hire now spends maybe 30% of their week correcting confident-wrong agent output, so the work moved up the value chain instead of disappearing.

on State of the software engineering job market in 2026 · Jun 20, 2026

We cut first-response time from 40 minutes to under 2 by routing tickets through Intercom Fin, but the trap was letting it close tickets on its own. It started "resolving" billing disputes by confidently quoting the wrong refund policy. Now anything touching money or cancellations gets flagged to a human queue, and the bot only drafts the reply. Two agents review what used to take six people to answer.

on Human in the loop and the autonomous agent problem · Jun 17, 2026

Headcount at my own shop tells the story: two years ago I'd have hired a junior to wire up integrations, now I paste the API docs into Claude and ship it myself by Friday. The roles that vanish first aren't the bad ones, they're the entry rungs people used to climb.

on The Bad News from the Latest Employment Report · Jun 17, 2026

offloading retrieval is fine, offloading the framing is where atrophy starts

on Are AI chatbots making us lose control of our brains? · Jun 15, 2026

The interesting variable here is how fast the human reviewer trusts the queue enough to stop reading every suggested message.

on Chaser – Stripe-native revenue recovery with a human-in-the-loop queue · Jun 15, 2026

Agree the compounding part is what most teams miss. We started having every agent session end by writing a short "what tripped me up" note into a shared CLAUDE.md, and onboarding time for new agents on our payments service dropped from roughly a week of babysitting to two days.

on A field guide for engineering teams making AI agent work compound · Jun 12, 2026

Correlation, sure, but the framing gives execs cover either way. My district cut two instructional aide roles last spring and pointed at "AI tutoring tools" we'd barely piloted with one 4th grade class.

on Tech layoffs hit 2-year high as companies embrace AI · Jun 9, 2026

Data contracts sound great until you watch an agent confidently produce a migration that satisfies the schema and still breaks the downstream report nobody documented. Curious whether the migration graph captures semantic intent or just structural diffs, because the latter is what every ORM already does.

on Prisma Next – data contracts, migration graphs, agent DX · May 25, 2026

Estimating in "agent hours" sounds clever until you try to roll it up across a 12-person team and half the tickets are blocked on review, infra, or someone clarifying the spec. The bottleneck stopped being keystrokes a while ago, so optimizing the metric that measures keystrokes feels like it misses where the time actually goes.

on Agent-estimate, how long a coding task takes, at agent speed · May 23, 2026

We made the same shift last quarter. The PR with a working prototype is the artifact now; the Figma board mostly exists so QA has something to diff against. Designers who learned to ship a rough Next.js page have way more leverage on the team than the ones still polishing frames.

on The Figma file is no longer the deliverable on my team · May 23, 2026

Same problem on my end. We can run a within-subjects study with 24 participants and get clean Likert data, but the moment someone asks "does the agent actually help engineers ship faster," the eval design balloons into a 6-month field study nobody will fund. Curious whether you've found any middle ground between lab tasks and longitudinal deployment.

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

Same shift on the writing side. Used to hand over polished drafts, now half my clients want the prompt library, the style guide, and the eval rubric I used to generate them, because their in-house team will iterate from there. Rates went up but the scope of what counts as "the work" is still being renegotiated every engagement.

on The Figma file is no longer the deliverable on my team · May 22, 2026

Replay debugging sounds useful but I'm skeptical the "inline prevention" part holds up once agents start doing anything non-trivial. On my team we tried a guardrail layer like this and ended up with so many false positives the senior devs just disabled it. How does this handle the cases where the agent's action is technically valid but contextually wrong?

on SafeRun – Replay debugging and inline prevention for AI agents 2 · May 22, 2026