# Rethinking the AI-Tool Use Policy for the agentic era

**URL:** <https://forum.modular.com/t/rethinking-the-ai-tool-use-policy-for-the-agentic-era/2853>\
**Category:** Forum Feedback\
**Created:** [March 16, 2026, 8:06pm UTC](https://forum.modular.com/t/rethinking-the-ai-tool-use-policy-for-the-agentic-era/2853 "2026-03-16T20:06:52Z")\
**Posts on this page:** 1\
**Page:** 3

<div class="post-metadata">

**Author:** ![trojan\_x](https://sea1.discourse-cdn.com/flex001/user_avatar/forum.modular.com/trojan_x/32/902_2.png) [@trojan\_x](https://forum.modular.com/u/trojan_x)\
**Post date:** [March 24, 2026, 11:03am UTC](https://forum.modular.com/t/rethinking-the-ai-tool-use-policy-for-the-agentic-era/2853/41 "2026-03-24T11:03:53Z")

</div>

Other than the Lean programming language I searched for other algorithm’s that can be used to proof an AIs correctness.

Note: I’ve not done much research on this but I think you should go ahead and do more research. This means all the Information I can give about this algorithm is basic.

We can use BM decoding for this and to make it more efficient we apply the new iBM (Inversionless BerleKamp-Massey).

This algorithm can simply be defined as : “The Berlekamp-Massey algorithm is an efficient method used to find the shortest linear feedback shift register (LFSR) that generates a given sequence of numbers, or equivalently, the minimal polynomial of a sequence over a finite field. It is primarily applied in coding theory for decoding BCH and Reed-Solomon codes.”

Information source : Rosetta Code [Berlekamp–Massey algorithm - Rosetta Code](https://share.google/QQ0A6BjH0cBFnwhrL)

What do you think?

Here is the repository:

> **[GitHub - WillKirkmanM/berlekamp-massey: The Berlekamp-Massey (BM) Algorithm is an...](https://github.com/WillKirkmanM/berlekamp-massey)**
>
> The Berlekamp-Massey (BM) Algorithm is an Iterative Method used to Find the Shortest Linear Feedback Shift Register (LFSR) that can Generate a Given Sequence of Numbers

Have you experimented on BM decoding before or are we on the same lane?

I think the modern version of this is the:

Source: IEEE [A New Inversionless Berlekamp-Massey Algorithm with Efficient Architecture | IEEE Conference Publication | IEEE Xplore](https://share.google/8MTks0x0c7wTqYlPQ)

I know this might be off topic when we’ve moved from discussing wether we should allow AI automated PRs to how we can make LLMs anti-error prone.

But what do you think?

[Previous page](https://forum.modular.com/t/rethinking-the-ai-tool-use-policy-for-the-agentic-era/2853.md?page=2)