vorticalbox

I’ve started doing a similar thing after reading a post on hn about manually applying the code so that you actually understand it.

I have done this for all my work this week and it works quite well.

For one it lets you actually query the LLM as to why, their plans give a high level not every single change and it allows you to correct it as you go and the plan will change.

have they fixed the file refresh issue?

Is it RLHF if there is no human feedback?

same things happens with npm and CVEs, most of the ones that bubble up aren't even called in my application so are no applicable.

go lang has a scanner [0] which only shows issues when the code actually calls the effected functions.

[0] https://pkg.go.dev/golang.org/x/vuln/cmd/govulncheck

that is not an model issue but a harness one, if you don't want it to have io to files or run command then simple don't give it access to the tools.

I actually use grok as my planner and then composer for coding like opusplan in Claude.

It seems to me that people have a dislike like for Elon then just blanket everything with the same label.

It might be called grok out it was trained by the composer team using a large chunk of their training data.

Cursor has two “modes” agent and ide.

In agent mode it’s horrible for editing files etc but better if you are juggling multiple chats

Putting the whole model in memory is far faster then swapping to disk.

I think cursor will likely do a grok fine tune rather than a kimi one for the next composer.

they noted in their blog post they didn't focus purely on coding for grok 4.5.

this is why i like golang these sorts of things just don't matter, go fmt move on with life.

I used to always have deep dread when having to use git but then I found [0]

[0] https://ohshitgit.com/

I encountered the same issue. Kimi 2.7 looked impressive on paper, but in practice, the code was so riddled with errors that I ended up using GPT-5.5 to fix it.

> I have to kick the hell out of its plans in ways that I don't think someone without a ton of real world experience could do

I think that’s the real point the article’s making. Unless you already know the ropes, you can’t tell if a plan is actually any good. I use LLMs all the time for work, but only because I get the system well enough to spot the nonsense or catch when a plan duplicates something that already exists. They really do keep rewriting the same functions over and over in the code.

in the same way claude and codex both use their paid models by default.

at least codex and grok are open source so we can see what is going on.

I always like testing out new things to put into my workflow. the first thing I will say is that its a welcome change to have an app 40 MB rather than over 400

i remeber reading in one of the release blog posts that that version was the "first that codex helped train"

thanks for the correction

The grok-cli is on github[0] there is nothing that I can see in the code that is activily looping ~/ and uploading everything.

My two guesses would be one the LLM decided it needed these files for the task or two the user simple asked grok to do it so they could post the tool calls on twitter.

[0] https://github.com/superagent-ai/grok-cli

why do people give these LLMs full access to everything and then complain when it does somethign stupid? that is what sandboxes are for.

Sure Is! I’m largely uninformed on the topic. Asking questions even stupid ones means you get to not be stupid after.

That’s good to know. This was one of my thoughts.

If the evs end up generating a tonne of material that we can’t do anything with that would be a short term gain.

Uninformed maybe, I’ve not looked into it as Evs not high my list of things to read about hence the questions that came to mind.

Maybe you can enlighten me.

My question that once you mined the materials for the car + battery, shipped that somewhere to be built, then shipped that to be sold.

Then you have the added waste battery once it not longer hold enough charge. Is it actually better for the environment?

If we add all this extra pull on the grid we need to generate the power for that too.

Well we don’t know but just look at figma, claude for it happened and then Claude design come out.

They knew exactly how developers worked from using figma as training data.

then why not just use threads/processes in python?

> There's a much cleaner way to do this - a shared helper

disagree, then you end up with something this this

function checkAll(target, conditions) { return Object.entries(conditions).every(([path, expected]) => { const value = path.split('.').reduce((o, k) => o?.[k], target); return typeof expected === 'function' ? expected(value, target) : value === expected; }); }

and const ok = checkAll({ user, account }, { 'user.isActive': true, 'user.isSuspended': false, 'account.status': 'open', 'user': u => u.hasPermission('read'), // predicate for the trickier bit });

how is that better?

if you child breaks the law and are under the age of an adult you as a parent are responsible, as a parent you are responsible to feed, wash, look after, keep dangerous stuff like medicine out of reach, if you live in a place that allows gun you have to keep that away too.

but suddenly because its a screen suddenly that responsibility to parent goes away?

i'm not trying to provoke I simple don't understand the other side of the argument.

do we know that Mythos is in fact a bigger model then opus?

They are offloading thinking to the AI too.

It then has to look over everything see how it connects together and then decide the best way to do something.

Giving it small and very focused plans when you already understand the system gets it done fast and cheap.