Why do you need them to provide a coding agent? Just use their model with any off the shelf coding agent. I happen to prefer Pi, but use whatever works for you.
I probably have an unfounded assumption that whatever coding agent they make will work really well with their models, better than external harnesses. I don't have a good sense for how all the model + harness combinations compare, nor any good way to compare them myself, but generally believe model companies train their models to work best with their own harness.
I've noticed that models have gotten less finicky with this over time. Harnesses don't need to be complex to get good coding performance from models, they just need to implement some sane primitives for code exploration and editing.
It is in the model's provider's interest for you to believe this because they get to lock you into their harness and inference. As models get better they will get better at using any harness, it comes down to how well the harness is actually engineered. I highly recommend you take an hour or two and check out Pi to either solidify or change your assumption. The harness is essentially just another developer tool and can be as opinionated, overly-engineered, minimal as anything else. I would think for DeepSeek, especially, they're efforts are much better spent researching how to make their LLM's better instead of working on engineering a harness that might get some marginal gain building it for their models.
> What's the best way to use it with Pi, OpenRouter?
I can't claim it's "the best"...
But the Pi.dev and OpenRouter combo is what I'm doing at home, and I love it.
Setup was easy, I can use /model to switch between any of the openrouter models and whatever I'm hosting locally via VLLM.
OpenCode Go is hard to beat, $10/mo ($5 first month) for up to $60 worth of tokens if used regularly. And no, you don't need to use opencode, they allow any software to use it.
That's not true. The LLM performance will degrade as the codebase gets messier as well. You get to a point where every fix breaks something else and you can't really make forward progress.
Yes, you might be able to get a bit further with a messy codebase just because the LLM won't complain and will just grind through fixing things, but eventually it will just start disabling failing tests instead of actually fixing things.
LLMs are originally trained to predict the next word in (mostly) human authored text.
Then they are fine tuned to follow instructions, and further reinforcement learning applied to make them behave in certain ways, be better at math and coding, etc.
They don't have any intrinsic motivation of their own, but they can try to parrot what they've seen in their training data.
So sometimes how you interact with them can affect how they interact, because they are following patterns they've seen in their source text.
However, a lot of folks use this to cargo cult particular prompting techniques, that might have seemed to work once but it can be hard to show that statistically they work better. Sometimes perturbing your prompt can help, sometimes you just needed to try again because you randomly hit the right path through the latent space.
I think your approach is probably a better one, for the most part trying to vary your prompt style is most likely to just affect the style of the output, so if you prefer a dry technical style, prompting it with one is the best way to get that out as well.
TL;DR With Flipper One, we're reimagining what a Linux cyberdeck can be — it's a huge
project. We're opening up the development process and asking the community for help.
Then later:
We're asking the community to help us polish RK3576 support so we can build a truly
open platform together. We'd be glad for any kind of contribution, not just code.
For example, maybe you can find a way to convince Rockchip to open up that last blob.
And:
Openness has always been our thing. With Flipper One, we want to go further — not
just open-source code, but an open development process. We're publishing our task
trackers, internal discussions, half-finished docs, and architectural debates. All
the messy stuff companies usually keep behind closed doors.
Then later:
We're also hiring a Developer Portal Manager — someone to act as a proxy between
our dev team and the community, help shape the Developer Portal, and engage with
contributors. Apply for the Developer Portal & Community Manager role.
Then they go into a lot more of the technical details of the process, with a few specific callouts of places they want help.
If you're into wireless work — auditing, monitoring, injection, mesh, anything —
we invite you to come test it with us: read the Wi-Fi Testing page on the
Developer Portal and help us decide whether this chipset is the right call,
or whether we should look elsewhere before we lock in the design.
I will say though: a lot of this has the feel of being LLM generated or "polished", which has the effect of making the brain kind of slide off of it. I know their team doesn't consist of native English speakers, so it's common for non-native speakers to use LLMs to try to polish their writing, but I find that the actual result is to make the writing have a just kind of bland personality that makes it harder to follow.
Tell me how I can easily determine the price from my IaC deployment as well.
Heck, I even have a hard time telling the price I pay on an account by account basis; because we have savings plans, those get charged against the root account and then I see $0 spent on EC2 in the individual account because it's all covered with a savings plan.
And when I'm putting together that IaC and trying to decide which new instance type to upgrade to, I have to dig through multiple confusing interfaces to figure out that what I want is to upgrade from m8a.4xlarge to c8a.8xlarge and how much that is going to cost me.
In cost explore, on the right hand side where filters appear, click the “more” button at the bottom then the first option that appears (which is “charge type” or something like that) and select “usage”. That will give you a view of what you’re using regardless of the savings plan, etc.
Thanks for the tip, but I tried that and I still see $0 for EC2-Instances, while if I look at Savings Plan coverage breakdown, I can see 100% of costs being covered by savings plans, broken down by instance family, but that view doesn't let me see things broken down by tag or any of the other ways you can view it in Cost Explorer.
If they want to be seen as responsible rather than opportunistic, then yeah, they should do a proper coordinated disclosure.
Sure, they have no legal obligation to disclose, but we all also have no legal obligation to buy their services. Blacklisting bad actors like this is the right move to discourage this kind of behavior.
they did a proper coordinated disclosure, following the industry standard 90+30 process. that is why the exploit dropped 30 days after the patch landed.
the kernel team should have communicated with their downstream about the importance of the patch. that is the kernel security team's responsibility -- and they are much better positioned to do that than crossing your fingers and hoping every reporter will contact every distro every single time there is a vulnerability.
there are very good reasons disclosure works this way, backed by a couple of decades of debate about it.
how many times it has to be said that it is impossible for linux kernel to communicate with anything but a minuscule portion of its downstream and _that_ has been done?
Who cares about how you are seen when you are selling 0day for big bucks? The bad actor makes more money than the 'legitimate' one without breaking any law. Punishing someone who didn't alert distros despite a patch being available encourages the company to simply find flaws and sell them for profit - it pays more to begin with.
If they want to take advantage of disclosure for marketing, they're either going to need to accept the norms around responsible disclosure, or they're going to need to accept how shirking those norms will come off. That's life in society. Sometimes it's annoying and sometimes it doesn't feel rational, but these norms have been negotiated throughout the history of our industry and are the way they are for reasons good and bad.
I just don't see the point in complaining about how shirking the norms of your industry will make you look irresponsible. I don't really care that they could have decided to sell the vulnerability instead. It isn't material.
It is absolutely not true that viable commercial vulnerability labs need to "accept the norms around responsible disclosure". There are no such norms. "Responsible disclosure" is an Orwellian term cooked up between @Stake and Microsoft and other large vendors to coerce researchers into synchronizing with vendor release schedules. It was fantastically successful at that, and it's worth pushing back on at every opportunity.
Tavis Ormandy dropped Zenbleed right onto Twitter. He's doing fine. You can blacklist him if you want; I imagine he's not going to notice.
Microsoft's policy is: "if you contact us with a vulnerability, you automatically agree to the terms of our responsible disclosure policy", which includes waiting 30 days after patch was created, and says nothing about how long that process takes.
There is actually no way to give them a friendly heads up, and then do your own thing. The only way not to be bound is by not sending them any notification at all...
You can email without agreeing to anything. But for a serious issue Microsoft would obviously try and track down who you are and what jurisdiction you are in.
> The Microsoft Bug Bounty Programs Terms and Conditions ("Terms") cover your participation in the Microsoft Bug Bounty Program (the "Program"). These Terms are between you and Microsoft Corporation ("Microsoft," "us" or "we"). By submitting any vulnerabilities to Microsoft or otherwise participating in the Program in any manner, you accept these Terms.
You said "There is actually no way to give them a friendly heads up, and then do your own thing. The only way not to be bound is by not sending them any notification at all..."
Maybe you're right. I just find it confusing. The language is all-encompassing, doesn't read opt-in to me if taken literally: "By submitting any vulnerabilities to Microsoft". And I found no other pages describing "report in such and such way to have these terms apply instead". But I always have problems with this stuff, perhaps taking it too seriously.
Obviously they can write whatever they want in their policy documents. The thing is, sometimes this is about larger sums of money, or someones reputation, which may or may not actually lead to steps. That is in contrast with whatever TOS/EULA in account signups for some service or whatever, this feels more serious. I've seen some people getting harried after publishing something that fell _outside_ the servicing boundaries. Getting tangled up in whatever is already a loss in my book, even if you "win" in the end.
Note that that policy is also where they set out the safe-harbor conditions, which, according to my read, is tied to the bounty policy and not RD/CVD policy. The RD/CVD page itself specifies no such thing, so I relate them.
I do not speak for MSFT, but last time I spoke with MSRC indeed they would be happy to receive your vulnerability report even if you did not wish to participate in any particular bug bounty program.
Those norms do not exist. Those are people asking companies to do stuff to benefit the person complaining for free, and many companies will not do that.
It seems to me you're unaware of them, but there are strong norms around disclosure. They've been discussed for decades. It is the expectation that vendors would be notified in a scenario like this.
No, there are users who want those to be norms. Qualified researchers happily sell substantive vulns to people who pay (Governments/Cellebrite and companies like that) enough to quell any complaint.
Which is again, irrelevant to the question of how disclosure works and what expectations there are around it because that is not disclosure and is not what was being discussed.
How does someone being incentivized to sell a vulnerability to a private organization over disclosing it publicly preserve a "high trust society"? Do you mean in the context of a "deceptively high-trust society"?
Those private actors aren't planning to sit around and hold onto these exploits they've horded forevermore, they're obviously paying for them so they can one day use them.
Gemma 4 31b was working ok for me; but it was consuming tons of memory on SWA checkpoints, I had to turn them way down, and as a 31b dense model is fairly slow on a Strix Halo. I did have a lot of tool calling issues on 26b-a4b, though.
My setup is a bit of a mess as I experiment with different ways of configuring and hosting local models. So at some point I was experimenting with the router server but stopped doing that, but some of my settings are still in models.ini while some are on the command line.
With the following as the relevant settings in models.ini (I actually have no idea if these settings are applied when not using the router server, it's been hard for me to figure out what settings are actually applied when using bot the command line and models.ini
[*]
jinja = true
seed = 3407
flash-attn = on
[unsloth/gemma-4-31B-it-GGUF:UD-Q8_K_XL]
temperature = 1.0
top_p = 0.95
top_k = 64
As my harness, I'm using pi, with a pretty vanilla config.
Anyhow, Gemms 4 31b worked in this config, but it was slow and RAM hungry. Since then, I've mostly moved to Qwen 3.6 35b-a3b because it's a lot faster.
I'm not actually doing anything useful with these yet, but I've used them for some experiments and Qwen 3.6 35b-a3b was capable of doing some pretty long mostly unsupervised agentic loops in my experimentation.
What do you mean, are they still making GPUs? This is a discrete GPU that has just recently been released, and it's one of the most popular GPUs in its class at the moment, due to 32 GiB of RAM for under $1000, which makes it great for LLM inference.
There was recent talk of them pulling back from the consumer segment, though obviously the leaks have also predicted Battlemage not being a thing so go figure: https://youtu.be/NYd2meJumyE?t=638 (timestamped)
That said, them not releasing a B770 in the consumer segment also sucks, since there are games and use cases that the B580 comes in a bit short for.
Generally the bottleneck is RAM throughput. Inference, in particular token generation, especially on a single user instance, is not all that computationally complex; you're doing some fairly simple calculations for each parameter, the time is dominated by just transferring each parameter from RAM to the cores. A 31B dense model like Gemma 4 has to transfer 31B parameters (at 16 bits per parameter for the full model, though on consumer hardware people generally run 4-8 bit quantizations) from RAM to the cores, that's a lot of memory transfer.
Prompt processing or parallel token generation can do a bit more work per memory transfer, as you can use the same weights for a few different calculations in parallel. But even still, memory bandwidth is a huge factor.
This is my biggest problem with the promises of agentic coding (well, there are an awful lot of problems, but this is the biggest one from an immediate practical perspective).
One the one hand, reviewing and micromaning everything it does is tedious and unrewarding. Unlike reviewing a colleague's code, you're never going to teach it anything; maybe you'll get some skills out of it if you finds something that comes up often enough it's worth writing a skill for. And this only gets you, at best, a slight speedup over writing it yourself, as you have to stay engaged and think about everything that's going on.
Or you can just let it grind away agentically and only test the final output. This allows you to get those huge gains at first, but it can easily just start accumulating more and more cruft and bad design decisions and hacks on top of hacks. And you increasingly don't know what it's doing or why, you're losing the skill of even being able to because you're not exercising it.
You're just building yourself a huge pile of technical debt. You might delete your prod database without realizing it. You might end up with an auth system that doesn't actually check the auth and so someone can just set a username of an admin in a cookie to log in. Or whatever; you have no idea, and even if the model gets it right 95% of the time, do you want to be periodically rolling a d20 and if you get a 1 you lose everything?
