HN user

encoderer

10,287 karma
Posts9
Comments3,107
View on HN

You might just be taking people too literally.

I've heard the same thing -- if they take a job they will lose money. What they really mean is that if they take a job (trade time for money), they will lose some of the pension they have already earned. This is a real economic loss (even if they might have a few more bucks at the end of the week) to say nothing of their lost time.

That’s a charitable read. Remember she also thought that New York would “lose millions” if Amazon put their hq2 there. Just because she’s in congress you can’t assume she understands how building a business works.

Taking your interpretation at face value I would add that yes we call it “capitalism” not “laborism”.

You're definitely burning more tokens with the back/forth and multi-step approach but assuming you swap who does the authoring from time to time you can definitely get the max out of each plan. Review doesn't use as many tokens.

I prefer having codex author plans and implement, and claude play reviwer. I do swap them from time to time and i have a lot of respect for claude 4.6 and 4.7 but for my domain I think codex does a better job with the authoring.

You can get a lot done with agentic programming without going "all in" on a gastown-like system, but I think there is a minimum viable setup:

1. an adversarial agent harness that uses one agent to create a plan and implement it, and another to review the plan and code-review each step.

2. an agentic validation suite -- a more flexible take on e2e testing.

3. some custom skills that explain how to use both of those flows.

With this in place you can formulate ideas in a chat session, produce planning artifacts, then use the adversarial system to implement the plans and the validation layer to get everything working e2e for human review.

There are a lot of tools you can use for these things but I chose to just build the tooling in the repo as I go.

Agent Skills 3 months ago

I adopted a couple of these, the api design and ui testing ones have been particularly helpful.

Such a long blog post with so little evidence.

How hard would it have been to create a test repo, make a branch with monolithic specs, make another branch with these scattered specs, run the same exact task against them both? Pick a few tasks. Run them a few times each. Analyze the results.

Yeah there are a lot of pathologies on both sides.

Democrats, for example, have loved to stoke the class war for the last 10+ years. Every week you hear a new take on how billionaires could solve homelessness or cancer or something if only they were taxed more. The left wing radicals are just as bad for America. Partisans hate it but the only true path out of this mess is centrism.

Totally, like I said right there, it is a great state!

But the governance is not great. That was actually the whole point I was making. And getting back to that: The getting things done is abysmal. Taxes are high. Spending is loose. No progress is made on things the state takes on as priorities (housing costs, high speed rail, homelessness). It's just not well managed. But from the "one side good, other side bad" POV it should be great, no pesky republicans to get in the way. I don't know if there's a lesson there but it's an intersting question to ask.

Where I live (California) there are basically no republicans in power. For a long time now. And we have more money and a better economic base than just about anywhere on the planet Earth.

So to all the partisans out there who are sure things would be better if "their side" had total control, I ask: what the hell is going on with California then?? We should be modeling the best governance in the country and even the world, but yet, our government is basically dysfunctional and our state is great despite it.

In my “repo os” we have an adversarial agent harness running gpt5.4 for plan and implementation and opus4.6 for review. This was the clear winner in the bake-off when 5.4 came out a couple months ago.

Re-ran the bake-off with 4.7 authoring and… gpt5.4 still clearly winning. Same skills, same prompts, same agents.md.

I know over a dozen 1-2 person SaaS, not including my own. Some of them have hired some help now but they are still more on the "lifestyle business" side. They are in many different spaces, and founders from around the world. I am not a big networker, but this is my niche and it's big enough that I just know a slice.