Dan Kubb @dkubb

github.com/dkubb Mission, BC, Canada Joined May 2007

Tweets

9K
Followers

2K
Following

3K
Likes

27K

Dan Kubb @dkubb

11 hours ago

@pvncher I haven't actually tried that specifically. Would it be two subagents, coordinating on some task provided by the main loop?

0 0 0 30 1

View Details

Dan Kubb @dkubb

2 days ago

@DavidKPiano @mattpocockuk DAGs of state machines is a really nice pattern for organizing work. Interaction Nets are even more elegant but they are not as intuitive.

0 0 2 94 0

View Details

Another tip is to ask it to do all the work in a YAML file, and to give each concept a unique id, and for it to have a dependencies attribute that is an array of those ids. Then it can write a program to sort the concepts and output them deterministically. The program can also check to make sure there are no loops and that everything forms a proper DAG. If it’s in a YAML other subagents can audit the file and point out references to concepts that are not listed as a dependency. You can also ask them to break the concepts down into smaller notes. I’ll often tell the LLM to use Zettelkasten style atomic notes, which will result in a single idea per note.

0 0 0 77 0

View Details

Dan Kubb @dkubb

2 days ago

@joseph_h_garvin I’ve had luck telling it to build up a DAG representing all the concepts, and dependencies between them. Then I tell it to topologically sort them, using relevance as a tie-breaker (weight relevant topics close together), and then display them in that order.

2 0 4 152 1

View Details

Dan Kubb @dkubb

2 days ago

@doomerpost @GreenTextRepost The world does not reward submission at all. School rewards submission. The world is not school. School is meant to produce workers that satisfy the needs of the elite.

1 0 4 445 0

View Details

Dan Kubb @dkubb

2 days ago

Coding is three things: 1. Thinking 2. Figuring out how to do something 3. Putting those thoughts into words That’s it. Those things are going to be valuable for a while longer. Let me put it this way: if you can’t do those things then you are dependent on someone/something else doing it for you; and they may not have your best interests at heart.

2 0 3 202 0

View Details

象牙山刘能 @disksing

4 days ago

在AI时代，每个程序员都活成了自己最讨厌的那种 team leader： 1. 半年不写一行代码，编程能力严重退化，还自诩"我也是搞技术的“ 2. 开会的时候把一线开发干的活全说成是自己的功劳，连自己的汇报材料都是让别人总结的 3. 瞎JB指挥，出了问题就甩锅说是下面没执行好 4. 不了解项目具体情况，只会提”单测加了没“、“性能还要优化”、“文档要沉淀一下”这类空洞的要求 5. PR 基本上看不懂，只会让别人先review了自己跟着点赞 6. 遇到 bug 了只知道无脑转发让别人查一下，自己负责来来回回转发消息 7. 还经常吐槽“今年的应届生水平越来越不行了”

222 534 3K 283K 724

View Details

Dan Kubb @dkubb

4 days ago

@JustDeezGuy @theodorvaryag @TheEduardoRFS That book is excellent. I don’t know why it’s not more popular. If you write SQL this book is a must read. It’s foundational.

0 0 2 32 0

View Details

Dan Kubb @dkubb

6 days ago

@charliermarsh “Gate” came out of nowhere. Never heard it before in the context of software. I can always tell who is AI-pilled when they say gate.

2 0 4 382 0

View Details

Dan Kubb @dkubb

6 days ago

@MoritzW42

0 0 2 661 0

View Details

Dan Kubb @dkubb

7 days ago

@justfielding Maybe the code was too enterprisey and it was like clippy “I see we are writing Java style Rust, let me help you with that!”

0 0 0 21 0

View Details

Dan Kubb @dkubb

7 days ago

I made GPT lose it's mind in my Codex App. Its last message (see screenshots) was to generate an image of a Java cheatsheet. I did not ask it to generate an image. I asked it to generate Rust code. And then it tried to compact over and over and even if I stop and restart or recompact it cannot break out of this loop.

1 0 3 319 1

View Details

Dan Kubb @dkubb

7 days ago

I hope someday @Apple will focus on writing software that is as good as its hardware. It'll be like Goku shedding his weights: youtube.com/shorts/KINrRCb…

1 0 2 142 0

View Details

Dan Kubb @dkubb

7 days ago

I could watch a whole series of lectures like this. I find this far more engaging than watching a slide-deck. The imperfect nature, and the natural pauses allows me to focus more on the material and absorb it rather than passively watching a polished presentation with picture-perfect slides.

1 1 3 785 0

View Details

Dan Kubb @dkubb

a week ago

To be fair, if someone is a 0.01x developer a 100x would bring their productivity to that of an average dev. I actually think this is what is happening in a lot of cases: non or really bad devs are finally producing a “normal” amount of code and are having the time of their lives.

1 0 2 125 0

View Details

Dan Kubb @dkubb

a week ago

One thing I do is set up a goal in a file, and then I tell Codex to use the goal file, and have Claude set to wake up every N minutes to review the commits and "steer" Codex by updating the goal file with flags it finds. Codex is also told about the arrangement and to re-check the goal file every time it completes a task. I literally do: /goal /tmp/goal.md This way I get Codex and Claude working together, with Claude watching Codex like a hawk. I'll also tell Claude to make sure Codex doesn't reward hack, and tell Codex to be wary of Claude too.

0 0 12 1K 7

View Details

Dan Kubb @dkubb

a week ago

There is definitely a change in cache usage. Around a week ago my 20x started draining 3-5x faster than before; with no change in prompt or project. I have been using goals with 5.5 low that drain 20-50% of my weekly quota in 24 hours. These are rust projects with long build times in between inference so it's not like it's hammering OpenAI servers or anything. There are multiple reports all across X confirming this. A 155h goal is a lot, but that doesn't mean it was all run consecutively. This is just how it reports things.

2 0 3 839 2

View Details

Dan Kubb @dkubb

a week ago

@zeeg @ThePrimeagen As I read that book a few times I was like “thats a very logical way to break that problem down and solve each sub problem in parallel”. I was not surprised af all to learn the author is a programmer.