Human oversight of agentic engineering
“Agentic engineering workflows still need human oversight, review and testing”
6 people · 5 agree · 1 disagree · 26 views · 2026-11-16 to 2026-11-20
Who stands where on the statement above, from their own words. A person's side is shown only where two separate readings of the quote agree; otherwise they are under "Side unclear". This page does not say who is right.
Agree 5 people
- ✓Theorem and Anthropic
Verification loops are the building block that lets you delegate more work to Claude.
If you have to focus on one test type, pick end-to-end.
The reason E2E tests are so effective is that they are difficult to fool–either the whole feature works or it doesn't.
Running the app end to end matters more than tests.
While unit tests are faster and less flaky, it’s possible to pass them and still break the feature.
Property libraries are a great way to keep agents honest by using edge cases that aren’t created with the same bias as the agent that created the original code.
Differential tests are particularly effective for large-scale migration use cases where functionality is expected to remain the same.
If your tests are flaky you’ll train the agent to run them again rather than fixing the failing code.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ✓Anthropic and OpenAI
We want engineers to spend their attention on decisions rather than reconstructing the agent’s work.
Updating an investigation skill or upgrading the model can make the agent better at diagnosing some cases and worse at others, so we need to test their impact before updating them.
our agents need monitoring just as much as the services they investigate.
Changes like these benefit from guidance from an engineer who knows the service and can weigh the tradeoffs.
- ✓xAI
We allow them to execute lower-risk tasks independently and require approval for sensitive actions such as issuing credits or refunds.
The people who know the job best remain responsible for defining how it gets done.
We start by teaching them how the job works, then have them observe us performing the work before gradually giving them responsibility for completing tasks on their own.
- ✓David Pogue
Instead, Pogue says there is more demand for software engineers today than there was two years ago. Why? Humans are still needed to review all the code that’s being created.
- ✓ankitg.me
the humans are still useful at defining how to iterate fast and what correctness looks like, and the agents are great to manage the exploration.
Disagree 1 person
- ✗Emilio Andere
AI is getting so capable that his engineers barely have to edit or review the code.
Side unclear 2 people
- ·Theorem and Anthropic
- Quote not reprinted: this page already repeats as much of this article as we allow.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- Quote not reprinted: this page already repeats as much of this article as we allow.
- ·Anthropic and OpenAI
Today, the bottleneck may be waiting for an engineer to review and merge the fix.
Week by week
- · 6: xAI ✓, Emilio Andere ✗, David Pogue ✓, Theorem and Anthropic ✓, Anthropic and OpenAI ✓, ankitg.me ✓
Built 2026-11-21 06:50 UTC from the news knowledge
graph by code. Every quote was checked against the source's own words; which side a quote takes is a model's reading,
made twice.