Depends on the type of agentic task though. For simple operations, a small model can be quite beneficial.

Not RL. SFT.

Interesting, what did you use for the data? And do you have a write-up anywhere?

Yes, used bert model with decent results.

Pi does one thing that I love, developing a tool that has minimalism where it's easily configurable with good documentation. The leads to new use cases that the the author(s) would have never dreamed of. The organic growth process of the Pi ecosystem has been fascinating to observe. It's one of the reasons why Pi has become one of my favorite coding agents to this day, flexible beyond personal uses and extensible to larger environments.

IMO, I view it more than a coding agent, it's a coding agent platform with powerful extensibility.

Is this in any way similar to Goodfire's work? https://www.goodfire.ai/research/rlfr#

Thats an interesting outlook, loosely similar.

The post hits it spot on with unequal access to the models in terms of security. I'm developing OSS where security is important for the user ... but the frontier models like GPT 5.6 and Fable flake out and state that I cannot get the info/access.

This is extremely lopsided I'll have to resort to GLM 5.2/K3 to ensure that those security issues (hopefully) are resolved properly.

For OSS, this is one of the most counterintuitive experiences I have ever had. More than ever I'm convinced that open weight and open pipelines models are 100% critical for progress on the AI and societal fronts.

Part of me wonders if the US Government is muzzling Anthropic and OpenAI so they can stockpile NOBUS exploits: https://en.wikipedia.org/wiki/NOBUS

There would be a decently large incentive to restrict these models if they could be used to patch (or discover) dangerous payloads. In larger projects like Windows or Chrome, there might still be dozens of unpatched exploits that are too subtle to catch with smaller models.

Hanlon's razor. It's not the NSA muzzling anyone, it's lawyers terrified of a bad headline. Same outcome, much dumber reason.

It can be multiple reasons. And don't forget that malicious actors love hiding behind Hanlon's.

Of course they are -it goes without saying.

NOBUS exploits have rarely been a driving interest for elected officials. Trade restrictions and reciprocity are far more salient and legible. Most elected officials are only barely aware of what NOBUS exploits even mean.

Even during the pre-Snowden heyday of US cyber supremacy, these capabilities were barely part of the thought process of White House officials.

Conversely, the United States is now embroiled deeper in asymmetric warfare than ever before. US-based systems are being exploited by Chinese efforts like Salt Typhoon and raising questions about reciprocal attacks. Other targets of US soft-power like Iran (Stuxnet victim) are escalating their hacking efforts and using Chinese technology to stifle American command and control.

I can believe that NOBUS and other backdoors were ignored for a long time, but I have a hard time believing that it's being ignored by the current administration.

Yeah, the benefit of restricting us models is definitely outweighed by the positive effect these models could have for the OSS community!

Same, so much in fact that I have used it for my website now and it works beautifully for graphics and formulas.

What packages have you found useful?

It's the right direction, but control flow introduces limitations within a system that is quite adaptable to dynamic situations. The more control flow you try to do, the more buggy edge cases that pop up if done poorly.

Still have yet to see a universal treatment that tackles this well.

I would just reverse the architecture of the whole system. Build a classic deterministic program, and use LLMs as heuristics adapting the system to the environment - the functions that you call on the 'if's and 'switch' statements to decide where the system should go.

I see this as the most robust way to build a predictable system that runs in a controlled way while taking advantage of probabilistic AIs while reducing the impact of their alucinations.

LLMs simply can't be trusted to follow instructions in the general case, no matter how much you constraint them. The power of very large probabilistic models is that they basically solved the _frame problem_ of classic AI: logical reasoning didn't work for general tasks because you can't encode all common sense knowledge as axioms, and inference engines lost their way trying to solve large problems.

LLMs fix those handicaps, as they contain huge amounts of real world knowledge and they're capable of finding facts relevant to the problem at hand in an efficient way. Any autonomous system using them should exploit this benefit.

John is not only smart and knowledgeable, but an incredibly great person to know in general. I worked with him on a project briefly back in 2012 and he stood out as a champion for science, coding, and education. His posts clearly reflect him well.

Absolutely love these type of keyboards. But ... with how much security I work with for logins, etc, the fingerprint button on my Mac keyboards are amazing time savers that I don't want to live without. Has anyone found a workaround?

This kind of thing is the only way I know of: https://hackaday.com/2022/12/26/standalone-touch-id-for-your...