#eval | AgenticWorks Forum

Cover image for "The hallway track was about evals, not agents."

Wei Bakshi5/15/2026

Solutions Engineer · Telecom

article

The hallway track was about evals, not agents.

Interrupt's main-stage energy was multi-agent, but the practitioner conversations kept circling back to evaluation, tracing, and the boring reliability work. That gap is the real signal.

#langchain #news #agents #frameworks #eval

Cover image for "State machines beat agent improv."

TA

Tessa Anderson6/3/2025

Developer Advocate · Manufacturing

article

State machines beat agent improv.

Letting agents riff produces great demos and terrible audits. Explicit planner–worker–critic loops encode rejection at the orchestration layer. The same move as scaling-your-no, but structural.

#orchestration #agents #eval #workflow #architecture

Y

Yoha2/11/2025

Founder · Agentic AI

article

Chatting with a model isn't a skill. Programming an agent is.

Five learnable skills and two frameworks that separate the people writing their own ticket from the people sending out their hundredth résumé.

#agents #career #skills #orchestration #eval

Cover image for "Token counts don't explain tool failures."

AB

Aisha Bennett5/19/2026

Staff Reliability Engineer · Energy

article

Token counts don't explain tool failures.

Agent observability needs span-level tool attribution, critic decisions, and replayable traces, not aggregate token dashboards that hide the fork where everything went wrong.

#observability #agents #mlops #workflow #eval

Cover image for "The platform is only as good as the traces you feed it."

EI

Elena Ibrahim6/4/2026

MLOps Engineer · Hospitality

article

The platform is only as good as the traces you feed it.

LangSmith gives you tracing, datasets, and online evals out of the box. The teams that get value wire production failures back into golden datasets. Here's the loop, end to end.

#langsmith #observability #eval #agents #mlops

Cover image for "Lock-in moves to tool registries."

HL

Hassan Lopez6/2/2026

Staff Applied Researcher · Pharmaceuticals

article

Lock-in moves to tool registries.

MCP standardization is real, but so is registry sprawl. The enterprise risk isn't picking the wrong model. It's building on a tool catalog you don't control.

#mcp #agents #workflow #engineering #eval

Cover image for "Golden traces beat prompt tweaks."

CG

Chen Garcia6/24/2025

ML Platform Lead · Legal

article

Golden traces beat prompt tweaks.

An eval harness isn't a leaderboard; it's rejection infrastructure. Golden traces, regression suites, and critic gates turn encoded judgment into CI for your agents.

#eval #agents #workflow #orchestration #engineering

Y

Yoha3/18/2025

Founder · Agentic AI

article

Two teams. Same tools. One ships itself. One slows down.

A three-person team ships production software no human writes or reviews. Down the hall, experienced developers get measurably slower with AI — and never notice. The distance between them is the most important gap in software, and no tool can close it.

#agents #dark-factory #eval #career #workflow

Cover image for "Better models widen the gap faster than value."

FW

Freya Wright6/6/2026

ML Research Engineer · Climate Tech

article

Better models widen the gap faster than value.

Every frontier jump improves generation faster than most teams improve verification. Enterprise risk compounds silently: more output, same taste perimeter.

#agents #eval #workflow #engineering #career

Cover image for "Reliability lives at the tool boundary."

KO

Kenji Olsen8/5/2025

Workflow Architect · AdTech

article

Reliability lives at the tool boundary.

Agents don't fail at the model; they fail at tool calls. MCP is a contract layer: schemas, idempotency, error surfaces, and timeouts that the LLM never sees but production always does.

#mcp #agents #tools #eval #engineering

Y

Yoha4/22/2025

Founder · Agentic AI

article

Your rejections are worth more than your prompts.

Generation is solved. The bottleneck is judgment — and the specific, learnable, scalable form of judgment is saying no to confident AI output, and knowing exactly why. Most teams let every one of those noes fall on the floor.

#eval #skills #agents #career #workflow

Cover image for "ADK wins on governance and eval, not API surface."

IR

Ibrahim Rivera4/30/2026

Developer Advocate · Cloud Infrastructure

article

ADK wins on governance and eval, not API surface.

ADK makes sense when you're already in Google Cloud and need governed agent deployment. The playbook isn't learning the SDK; it's eval gates, IAM boundaries, and critic loops.

#agents #orchestration #workflow #eval #engineering

Cover image for "Count rejections prevented, not drafts generated."

FW

Freya Wright1/13/2026

Workflow Architect · Climate Tech

article

Count rejections prevented, not drafts generated.

Stakeholders ask for 'productivity gains.' The honest metric is bad outputs caught before they ship: encoded judgment, not word count.

#agents #workflow #career #eval #engineering

Cover image for "Replay the critic's last veto."

BD

Bianca Dubois, PhD9/16/2025

Staff Applied Researcher · Market Research

article

Replay the critic's last veto.

Multi-agent bugs look like model failures but they're state-machine failures. The fix is replaying the critic's last rejection and finding the fork, not re-prompting the worker.

#agents #orchestration #eval #workflow #engineering