Mooneeb

A simple guy thinking about engineering in the agentic age. Also comics, sci-fi, games and other things nobody asked me to build. Open to work.

[Mooneeb @Mooneeb]

@stfusaadd @ufffencive Afaik - and I could be wrong - they probably have methods in place to make sure the quality of work meets the needed standard. Like a list of things to check to make sure work is good. And they have a lot of hierarchy so they have individuals who are responsible for validation.

[Mooneeb @Mooneeb]

Humans will eventually have Agents explain codebases to them. That doesn't mean humans shouldn't and won't read code - they will, and the code needs to be readable. But you don't need comments for that, and neither do Agents. And the todos etc. Should exist in some tracker. 4/4

[Mooneeb @Mooneeb]

Agreeeeeeeed 100%.

From what I've seen across multiple codebases and this is my opinion as well, comments existed as a means of communication to other humans in the future. A nit here. A ToDo there. A way to let people know there was an idea that was cool but not implemented 1/

[lauren @poteto]

ban and delete most if not all code comments!

agents love copying what already exists. a seemingly harmless comment explaining some temporary workaround is one of the worst viruses you can have in your codebase. like a weed it'll spread quickly as a tool for agents to quickly get their job done. before you know it, your codebase is overrun with hacks and stale comments. just one more workaround bro

you can use pstack's /no-comments skill as a weedkiller. it'll seek out useless comments while keeping the ones that actually should exist (like explaining a browser quirk), and figure out how to solve them on a first principles basis

here's a prompt you can try asking @bot:

create a project agent that uses /poteto-mode to spawn a /swarm to use /no-comments and identify useless code comments we can prune in my repo. come back with a high level plan of which of these workarounds would be best solved by refactoring and which we can straight up delete

https://t.co/nqeRdY62ee

[Mooneeb @Mooneeb]

code without comments. A developer should know what the file does without needing extra assistance.

This is moreso true with agents now more than it ever was because with agents, there should rarely (if ever) be a situation where you're having to create nits and todos. 3/

[Mooneeb @Mooneeb]

@thdxr True but there's nothing. Wrong in having an opinion. I could be completely misreading the suble hint at the discussion topic, but if I'm not, wouldn't you say the way you work dictate your opinion on the subject matter

[Mooneeb @Mooneeb]

@theo And it's definitely something to do with my lack of knowledge around T3code, but could you please help me understand how to best utilise T3code in such a scenario? I'd love if there were examples of your or other people's workflows I could learn from as well. 3/3

[Mooneeb @Mooneeb]

@theo Hey Theo, long time Twitter stalker, big fan. I've been using T3code these past few weeks trying it out. Love the interface. But the concern that keeps popping in my head is how to use it to maximize the context window because I subscribe to Matt Pocock's style of thinking 1/

[Mooneeb @Mooneeb]

@theo That every agent has a smart zone and a dumb zone, and the smart zone is a small section of the overall context window ~200k/1M tokens. When I use T3, I find that it's great when I don't worry about that context window (I don't much with codex) but I get paranoid with Claude 2/

[Mooneeb @Mooneeb]

@poteto Things are better and they'll continue getting better, but isn't "AI slop" something you are worried about still? Because avoiding it requires some amount of maintainence - or do you believe that we'll reach such a state that any such "slop" will be easy to fix in the future

[Mooneeb @Mooneeb]

@poteto Like, I understand in the future (6/12 months) things may change and we should build towards that, but isnt the idea of "software factory" relying upon having deterministic, human-trusted verification tests that agents can build upon and use as their own foundation & reduce slop?

[Mooneeb @Mooneeb]

@wagslane Bash script* and lot of slop*

[Mooneeb @Mooneeb]

@wagslane +1 Hyperbolic to say but half the internet is propped up by that one comment less nah script that someone wrote back in 2000s. I do think there is a lot of skill right now but imo its cause we're transitioning to the use of this constantly developing tech. Things will get better

[Mooneeb @Mooneeb]

People who're using @t3dotcodes how do you keep check on your agent's context window so that it doesn't exceed a certain threshold? Using Claude and codex both? Do you just set a lower autocompact setting?

[Mooneeb @Mooneeb]

@daadapastha Curate your linked in and social media feeds such that you get recommended job openings. You'll start coming across people like Andrew Stetsenko who devote time to curating job opportunities for people like us. And then it's just about applying and praying.

[Mooneeb @Mooneeb]

@poteto @bot Curious on your thoughts. But if overused, won't the domain fall culprit to spam detection?

[Mooneeb @Mooneeb]

@mattpocockuk Recommendations on books to read to improve as a software engineer and more importantly as an "AI Engineer"? 🫣

[Mooneeb @Mooneeb]

@listelia Interested. I have multiple years of experience working with and building AI powered systems while also using AI tools to improve coding efficiency and productivity. I can share further if required