scwoodal

Can you share examples, links to Github of this approach? I'd like to learn.

Enforcing structured outputs from LLMs is not the same thing as using linters, formatters, static analysis to control how an agent writes code.

> I'd rather see a coding agent that can follow steps in a plan file to a T while following guardrails and adhering to the proper coding conventions in the human reviewed spec.

Guardrails/conventions should be enforced in linters, formatters, static analysis tooling; not specs/prompts.

My parents have been religiously listening to the likes of Rush Limbaugh since the early 90's on a daily basis. I recall listening to it on our drive into work together.

30+ years of hearing whatever the tv/radio host du jour says, without critical thinking, taking it at face value, 8+ hours every day. In the car to work, radio on while working, drive home, and then turn the TV on till bed time.

Then take a drive through rural America and see that education isn't an important pillar.

I went to college (neither of my parents did) and made it out with a different perspective, but my brother did not. He peddles in conspiracy theories and doesn't believe humans landed on the moon. My father regrets sending me to college.

Pretty good as others have mentioned. Even more so when one starts adding packages like credo, ex_slop, and boundary to run deterministically and steer the LLM.

Don’t like a pattern that LLM introduced? Write a custom credo rule and the problem doesn’t come back.

AI makes it trivial to narrow down who this company might have been despite the anonymized details. Mostly a heads up for future posters who want to truly avoid it.

Except after 4 gallons it might as well be pure oil, mucking everything up.

> NCAR’s climate research may not be the only thing driving the White House attack. It is widely believed to also be part of a campaign of political retribution waged against Colorado for its conviction and imprisonment of Tina Peters, a former county clerk who breached election security systems in a scheme to find proof of fraud in the 2020 presidential election. The state has included the attack on NCAR in a lawsuit against the Trump administration.

Federal policy by wounded ego is still wounded ego.

My Claude sub isn’t unlimited.

I didn’t set the limits on the plan; change those if it’s a problem, not irritate your customer base.

So I’m not allowed to use the $20 plan and max out its limits?

I’m not subletting my sub to anyone. I’m the only one using the third party harness.

I’m using their own SDK in my own CLI tool.

Why does it matter to Anthropic if my $200 plan usage is coming from Claude Code or a third party?

Doesn’t both count towards my usage limits the same?

This is correct. It just takes one person to review it and you’re good to go.

There’s also a service that rates your grammar/clarity and you have to be above a certain score.