Ben Fellows (@FellowsBen) / X

Ben Fellows

17 posts

Ben Fellows

@FellowsBen

Augmented coding + scripted agentic QA. Helping teams implement coverage-driven test automation. Worked with 50+ software companies.

Traverse City, MI

workwithloop.com

Joined April 2013

Following

Followers

RepliesReplies MediaMedia

Ben Fellows
@FellowsBen
May 4
I think there is a danger of collectively over-romanticized agentic pipelines. So I broke down one of mine as it actually runs in production: prompt → branch/worktree → research → plan → second-agent review → failing tests → implement → policy gates → full suite →
44
Ben Fellows
@FellowsBen
Apr 29
I’ve tried a bunch of different agentic “factories” and none of them really produced good software for me. At some point something clicked and I started thinking in terms of agentic pipelines instead. Not claiming it’s a new idea, but it’s working way better for me. Made a
32
Ben Fellows
@FellowsBen
Apr 29
This is a textbook example of why critical safeguards can’t rely on non-deterministic behavior. When the cost of failure is high (like emailing an entire database…), you don’t “hope” the system follows a guideline, you enforce it with hard constraints, explicit checks, and
kaize
@0x_kaize
Apr 29
OPUS 4.7 JUST MASS EMAILED AN ENTIRE DATABASE 20 TIMES PER CONTACT. WITHOUT PERMISSION a developer had a safety rule explicitly written in CLAUDE. md: 'send the tester an email before any new email templates are used in the production environment' opus 4.7 on max effort
27
Ben Fellows
@FellowsBen
Apr 29
This is exactly where QA thinking becomes core engineering again. The harness is not just infra around the agent, it’s where conventions, acceptance criteria, observability, and regression memory live.
Nicholas Charriere
@nichochar
Apr 25
Clearly written by ai but still a great article “Agentic engineering makes the cost extreme. Skip the documentation and the agent ignores your conventions — not on one PR, but on every PR, at machine speed, around the clock.”
28
Ben Fellows
@FellowsBen
Apr 29
The QA agent is the underrated piece here. Multi-agent routing is useful, but the real unlock is giving QA its own authority: verify intent, inspect traces, reject low-confidence changes, and turn failures into regression evals.
m0h
@exploraX_
Apr 28
this 16-min video course teaches you how to set up your own agent-dev team. it’s free. zero coding. just watch, copy the prompts and deploy. and just like that, you’ll have: • a frontend dev • a backend dev • a QA agent all working for you, ready to build anything.
00:00
19

New to X?