schaefer

Correct. I am only working on building my own harnesses.

I don't know if you'll be interested, but there is an excellent book on the topic: Build an AI Agent (From Scratch), by Jungjun Hur and Younghee Song (Manning Publication).

By chapter four you have a minimal working codebase in python. Later chapters add out features like tool calling, planning, and reflection.

Thanks for the suggestion, but that’s not a fit for me.

I’m writing my own harness to learn. And I’m writing it in zig.

But, I’ll absolutely add maki.sh to my reading list. So, that’s a win.

> "What kind of hardware... "

At home, I mostly use openrouter.

A few bucks will realistically last you forever. I also bought a DGX Spark at launch. I've learned a ton. but it's not usually my daily driver. I do run overnight tasks on it, like "read this book and write up notes on each chapter..."

In retrospect, given the dirt cheap prices of apis and or renting hardware, buying hardware for inference was likely a mistake (but not enough of a mistake to sell it now).

Pi.dev at home.

And I’m slowly writing my own. But it’s garbage at the moment.

I just looked up a video for this game, and looks like a really interesting take on user interface in VR. Thanks for the head's up!

>> Not sure there's a strong use case…

The author’s discussion of their personal use case starts on the seventh sentence in TFA. And frankly, it sounds like a pretty solid use case:

“Being able to run things locally and mess around with RL, quantisation and fine-tuning on a laptop would be sick. I already publish my own GGUF quants on Hugging Face, with over 160,000 downloads in the past month.”

I agree with your skepticism.

3.8k is less than the preorder cost of the DGX Spark over a year ago (Oct 14th, 2025).

This demo took place roughly 30 hours after Starlink betrayed the Ukraine by dropping support for Ukrainian high speed intercepts and drones [1].

I don't see how we can discuss this demo without acknowledging that fact as well.

[1]: https://apnews.com/article/russia-ukraine-war-zelenskyy-burn...

this is not what *decision models* do.

I would like to start seeing bar graphs for benchmark results to add a second y axis to show token spend or cost in dollars.

I guess what I'm saying is performance is fine... but what I really want to see is performance/dollar.

In my state, early voting in the US midterm elections start on October 17. That's just eight days from now.

Thank goodness.

This is a great recommendation. And considering the code base is in assembly, I'll gladly welcome a walkthrough.

thank you!

Is any of the original Apollo code available to read?

but to do accurate billing, they would have to tell us when our usage triggers a guard rail and causes our query to be rerouted or denied.

And if they do that accurately, there's no stopping someone from mapping the guard rails.

Clearly that outcome is more uncomfortable than what we have now, which is enough people seem to be okay with vague billing.

any chance you have a public git repo for that personal harness of yours? curious minds want to know.

did I miss something?

There is no author attribution to this webpage?

I feel strongly that I can't take a community challenge seriously if the author can't bother with an author attribution.

Frag,

I am engaging in this from a "Logic 101" perspective. I saw a false premise used in an argument and I only wanted to point that out.

I have no actual position on cars brands or apple computers or any of that.

In most countries without 100% tariffs and outright bans Toyota's sales are in decline because of competition from BYD.

For example, sales for the first half of 2026 vs first half of 2025 in Australia show:

Toyota is down 20% BYD is up 68%

Let me be more clear. The activity you're advocating is illegal:

"possessing, discharging, or igniting explosives or explosive material, including binary explosive targets is banned on BLM land in nevada under Federal Land Policy and Management Act (43 U.S.C. 1701 et seq.) and 43 CFR 9212.0-6."

The phone number to report illegal activity on BLM land is: 1-800-637-9152.

what a funny Freudian slip for skew.

> Some rich person could just setup a contest with awards and see who shows up, it could be held out on one of those dry lakebeds in Nevada.

As a resident of Nevada - HELL NO. That's BLM land. Not an amateure weapons test site.

If "some rich person" wants to get involved in crowdsourcing weapons research, please do that on private land... so everyone will know exactly who to sue should things go sideways.

I already own one - I've been reading with crosspoint.

I hadn't heard about microslate. Thank you.

Yes, I meant slow down.

With effort, the effect can persist. But you have to be willing to make changes.

Go on a ten day meditation retreat.

Time will hit the breaks, hard.

Right?

Between $18k-40k to run a benchmark.

Somewhere around 2004 I had an experience at the black hole that still amuses me.

I was digging through the store when I found a full pallet of computer punch cards. As a computer science undergrad, I thought, “ how cool, I want a bunch to use as bookmarks, or whatever”.

I went to haggle with Ed, and he outright refused to sell me any. He said he wouldn’t break up the pallet. I had to buy all or none…

That’s awesome. I’ve seen the book around, but didn’t know the connection to the works of Borges.