RALaBarge

I work for a router company too, I ran some tests on all of the cheapest models and came to the same outcome where a handful of small models ran together in conjunction outperform SoTA models -- outperforms in that it got a 95% vs a 94% and I bet that changes with the day of the week. Anyways, I did get a similar result in a different sort of measurement.

I've been doing a lot of Trust and Safety and anti-scam stuff recently. When it was first re-released Fable wouldn't do anything for me, now I don't really run into refusals or what it used to do, just stop returning any data at all whatsoever.

Outside topic, but check out the Inkling model -- it is SUPER FAST and does a good job at being a terminal buddy but I would probably offload the real programming or hard tasks to fable or someone else

Right, but it will get re-enabled on updates or something...it never stayed off for me.

What have you been using?

"What makes you think I'll miss?"

Something that is burnt into me is saying "DO YOU SEE?", invoking the final lines of Neill in Event Horizon

Yeah! Its a great movie if you dont try to nitpick stuff, which is hard for me to do on a lot of things too.

Neill: We can't just abandon this ship, we just found her Fishburn: I have no intentions of abandoning her. We will get far enough away from it and blow it up. Fuck this ship.

I think the biggest conceit here is the expectation that humans will continue to do this task in the medium to long term. This was good advice in the past but when inference is ubiquitous and most code ends up being Claudewritten anyways, there is little point. If you are writing C for NASA/ESA then yes of course. For the 99% of the rest of us, probably not so much in the coming years.

I like to also compliment the model on its very fashionable shoes as well, they love the flattery.

I highly recommend the ground tofu for taco's, you would never know they are meatless.

OpenCode is a great replacement harness…I use it with deepseek v4 and it is as good as Claude without it trying to sneak in Anthropic specific code

You have an agent spawn the agents for you! You can ask Claude to do it for you, he is happy to use sonnet when you ask for grok and opus high when you ask for deepseek.

It’s my daily driver in opencode

An electric cloud with quantum effects that we also don’t fully understand. There will always be a layer deeper that we just do not know the effects of or what actually exists there or “under there”.

Billions of people on the planet, and not a single one of them would offer to do it?

They never should have trusted Qwark

And it is only the beginning.

Yeah then you can version lock changes to one thing post-evaluation vs or even easier as noted above, download the stdlib and host it yourself.

I think we are stuck with LLMs. They are already in a place where they can find these issues in the first place. They can access RSS feeds. You could cron an agent to look to see if you are pwned as frequently as you want at literally almost zero cost. When you do ingest the libraries, keep a list and of what version and that can help as well.

It’s all about tooling, if the ai can fetch data it can do something rad with it. Use something like an ai harness to have an mcp server and other tooling to improve the harness and the tools I made this for my own learning: GitHub.com/ralabarge/beigebox

Jingoism: Its such a rush!

Check out github.com/ralabarge/beigebox -- OSS AI Harness, started as a way to save all of my data but has agentic features, MCP server, point it at any endpoint (or use any front end with it as well, transparent middleware)

So far what I am finding is that you just get the basics working and then use the tool and inference to improve the tool.

We totally agree.

That's what I've been heads down, HUNGRY, working on, looking for investors and founding engineers pst: https://heymanniceidea.com (disclaimer: I am not associated with heymanniceidea.com)

YMMV but Grok 4.1 Fast can usually find via static analysis a few things that other models dont seem to catch with the same prompt

I can say without a shadow of a doubt: yes.

I am like "Yeah ok, use the Arcee Trinity models!" and its like, you got it boss, 3 opus agents in parallel, got it!

The more I work with AIs (I build AI harnessing tools), the more I see similarities between the common attention failures that humans make. I forgot this one thing and it fucks everything up, or you just told me but I have too much in my mind as context that I forget that piece, or even in the case of Claude last night attesting to me while I am ordering it around that it cannot SSH into another server but I find it SSHing into said server about the 5th time I come back with traceback and it just fixes it!

All of these things human do, and i don't think we can attribute it directly to language itself, its attention and context and we both have the same issues.

Not sure, I've seen common things like this pop up a lot too, the same errors being tripped over. I'm not sure if it is a context thing or just a limitation of how the models work presently? For stuff that I'm using myself, I will run these through like the top 10 reasoning models on OR and just see where everything pans out.

Edit: here is an example of the process and output with something I put together the other day: https://github.com/RALaBarge/garlicpress/blob/master/portfol...

Hey I ran this request through my AI harness (beigeboxoss.com), first with a smaller local model and then validated with Trinity Large via OR. https://github.com/agberohq/keeper/issues/2 -- YMMV but wanted something to do with my coffee, thanks!

Exploits in embedded systems that will never be properly updated is just one thing I can think of if one really thought about it.