Kamal Mustafa

Software dev, technologist. Solving business problem with tech. Run #belajarpythonmy on Telegram.

[Kamal Mustafa @k4ml]

@lomo_tech Mostly useless utk coding agent mcm Claude code or opencode. 40GB GPU, Nvidia.

[Kamal Mustafa @k4ml]

Finally!

https://t.co/PfKiPI1XfO

[Kamal Mustafa @k4ml]

One github would be killer feature - `copilot --web`. Make it accesible from https://t.co/eLPFPeeCmv but all executions on your own machine. Rn the only sharing options you have is /share which only post to gist or make the agent use gh cli to update issues/pr.

[Kamal Mustafa @k4ml]

@khairilazizee Yg free banyak, opencode, kilo, amp, Gemini cli, mistral-vibe, chatgpt codex pun ada 1 month free trial.

[Kamal Mustafa @k4ml]

... but now you can't scroll up using mouse anymore. To accomodate they also introduce --no-alt-screen in case you prefer the old screen flickering behavior. But then they removed so now you have to choose - copy-paste but no mouse or keep mouse support but no copy-paste.

[Kamal Mustafa @k4ml]

github copilot cli has problem with screen flickering once the session grew to certain length. It's the worst among coding agent that I'm using so far. So they fixed and that comes along with broken copy-paste on ssh, same as opencode. They fixed that as well with --no-mouse ...

[Kamal Mustafa @k4ml]

@afifsohaili nothing beat good code for AI to start with and immitate. When that not enough, that's where skills supposed to help.

[Kamal Mustafa @k4ml]

quickly vibe code waktu solat app. Started with claude and then refining with codex cli. Work offline. The calculated time might be off by few minutes but to me it's good enough for personal usage.

https://t.co/nYDkGXiGWK

[Kamal Mustafa @k4ml]

We have great past civilization but disconnected, partly through colonial reset that last few hundred years. We remember the glory but not the mechanics. Without mechanics, history becomes pride, not power.

[Kamal Mustafa @k4ml]

but it can be flipped to other side as well.

Why hire seniors when juniors + AI can do the same tasks at much cheaper cost? And this solve the talent pipeline above as juniors can eventually grow into seniors instead of retiring.

[Kamal Mustafa @k4ml]

Where's the seniors who can supervise the hallucinating LLMs gonna come from? The moon?

Juniors not needed because AI can do everything at that level. But seniors need to come from somewhere, i.e juniors. Can't find satisfying answer to this yet.

[BURKOV @burkov]

This is the truth: modern agentic coder AI is better at everything compared to an entry-level coder.

**At everything.**

This means that hiring an entry-level coder doesn't make any practical sense, unless the organization wants to invest 5 to 10 years into someone learning to code by hand at a level of a mid- or senior-level coder (which is currently still needed to oversee agentic coders).

No leader in modern business would invest in an employee education for more than a quarter or, at the very least, a year.

No one.

Unless we reintroduce slavery, where the slave worker is required to stay with the employer who invested in their education until the debt is paid in full. Which is, of course, ridiculous to imagine, right?

[Kamal Mustafa @k4ml]

when running openclaw in container we shouldn't use the systemd service, starting browser will fail. I try with xfbdev but no use. For now using xfreerdp to login to the container and start openclaw gateway from console.

[Kamal Mustafa @k4ml]

I'm not wrong then for personal project's AGENTS.md is only 5 lines.

[Addy Osmani @addyosmani]

Tip: Be careful with /init. A good mental model is to treat AGENTS(.md) as a living list of codebase smells you haven't fixed yet rather than a permanent configuration.

Auto-generated AGENTS(.md) files hurt agent performance and inflate costs because they duplicate what agents can already discover. Human-written files help only when they contain non-discoverable information - tooling gotchas, non-obvidous conventions, landmines. Every other line is noise.

Beyond what to put in it, there's a structural problem worth naming: a single AGENTS(.md) at the root of your repo isn't sufficient for any codebase of real complexity.

What you actually need is a hierarchy of AGENTS(.md) files - placed at the relevant directory or module level - automatically maintained so that each agent gets context scoped precisely to the code it's working in, rather than a monolithic file that conflates concerns across the entire project.

[Kamal Mustafa @k4ml]

@azrulrhm @openclaw been using termux on phone to connect to incus containers in my laptop running all those coding agent - copilot cli, gemini, opencode, mistral-vibe and kilo cli.

[Kamal Mustafa @k4ml]

That's it. Claude code implemented exactly this. @github what you're doing?

[Kamal Mustafa @k4ml]

One github would be killer feature - `copilot --web`. Make it accesible from https://t.co/eLPFPeeCmv but all executions on your own machine. Rn the only sharing options you have is /share which only post to gist or make the agent use gh cli to update issues/pr.

[Kamal Mustafa @k4ml]

One github would be killer feature - `copilot --web`. Make it accesible from https://t.co/eLPFPeeCmv but all executions on your own machine. Rn the only sharing options you have is /share which only post to gist or make the agent use gh cli to update issues/pr.

[Kamal Mustafa @k4ml]

Their answer to google cloud run I guess. I can't even remember when was the last time I'm looking for aws equivalent of cloud run, like many years ago.

[AJ Stuyvenberg @astuyve]

Perhaps my favorite new AWS launch is ECS Express Mode.

You need a container image and some credentials and... that's it!

you get an autoscaling container with sane defaults

But the coolest part is that you can easily create a public or private application, with no extra work

[Kamal Mustafa @k4ml]

Running podman inside incus container.

https://t.co/VItXGwubdv

[Kamal Mustafa @k4ml]

... will max out the cpu. It's ironic frontend tooling now require more resources than backend. With push based deployment, we can run the build on more powerful machine and just push the build output to deploy. Something we did with https://t.co/mVS3QtSGSq in the past.

[Kamal Mustafa @k4ml]

have to think again on push vs pull based deployment. I'm long believer of push but pragmatic need make us switch to pull for most of our projects now. But there's one instance where we need to deploy on lowend vps because it just a small wagtail site. `npm run prod` however ...