d4rkp4ttern

7 ms·

31.
by d4rkp4ttern 3mo ago
Me: Ok so I'm going on a walk. I'll dump a link to a Hacker News discussion about an article. You have to read the article and the discussion and walk me thru all the interesting details, nugget by nugget,
32.
by d4rkp4ttern 3mo ago
When notebookLM was new, it was interesting to listen to the podcasts. Then the novelty wore off, and I wanted something where I can interact with the podcasters but it was janky as hell. My current “audio-learning” hack is ChatGPT Live whi
33.
by d4rkp4ttern 3mo ago
I've been looking around for skills that help with improving AI prose, and found these: (1) Avoid AI Writing - https://github.com/conorbronsdon/avoid-ai-writing (there is a similar "humanizer" skill that
34.
by d4rkp4ttern 3mo ago
I too used to be a skeptic of the whole "agent-swarms" hype. But I do find some patterns useful with ultracode: - parallelism: "if it makes sense, use ultracode to do this thoroughly" Claude decides if this is the type
35.
by d4rkp4ttern 3mo ago
I’m curious why many tech bloggers choose Substack, when it’s almost trivial to set up your own blog with infinitely more style control, and math/syntax-highlighting etc.
36.
by d4rkp4ttern 3mo ago
My understanding is that in the generated JS code, the subagents are invoked as headless Claude, equivalent to Claude -p, or the agents SDK.
37.
by d4rkp4ttern 3mo ago
> single line addition to prompt to use subagents proactively. This misses an important detail. In Claude Code [1], ultracode suggests the agent create a JavaScript code to deterministically orchestrate sub agents. This is different from
38.
by d4rkp4ttern 3mo ago
That’s a good point
39.
by d4rkp4ttern 4mo ago
100% Also good to note that Cherny and Steipete have said in interviews that they keep it simple and do not use any of these shenanigans.
40.
by d4rkp4ttern 4mo ago
I didn’t know you could get the “Chat-GPT-5.5 Pro” (the one that’s been solving Erdos problems) inside codex-cli, or maybe I misunderstood?
41.
by d4rkp4ttern 4mo ago
Ok I’ll ask since nobody else has — are they not giving their devs a Claude code max or Codex Pro subscription? If so, why is token cost approaching billions? And if not, why not?
42.
by d4rkp4ttern 4mo ago
When I tried handy a few months ago the transcription had stuttering issues, and I later discovered Hex and it was even faster and did not have those issues, so I stuck with it. This was with the same underlying Parakeet model, I think the
43.
by d4rkp4ttern 4mo ago
Nice collection! Do you know if any of those can do meeting transcription with turn detection with consistent speaker ID ? Otter killers in other words.
44.
by d4rkp4ttern 4mo ago
My current fav STT is Hex, MacOS only: https://github.com/kitlangton/Hex Near-instant accurate-enough transcription using local Parakeet V3. Handy is multi-platform, also great: https://github.com/cjpa
45.
by d4rkp4ttern 4mo ago
> It’s available today in beta for Claude Enterprise and Team customers. You’re welcome.
46.
by d4rkp4ttern 4mo ago
> The engine decomposes the traces to understand common failure modes across harness executions and produces a report with its findings. What are some examples of these common failure modes?
47.
by d4rkp4ttern 4mo ago
Is this a generally well known thing?
48.
by d4rkp4ttern 4mo ago
TechCrunch articles should be ignored into oblivion.
49.
by d4rkp4ttern 4mo ago
Thanks, not sure what went wrong, but fixed it now
50.
by d4rkp4ttern 4mo ago
I sometimes have claude or the coding-agent quiz me, Socratic style. It persists in asking questions at deeper levels until you arrive at the answer yourself. This forces you to think hard about a problem, and this effort helps with underst
51.
by d4rkp4ttern 4mo ago
You're thinking of the legendary IIT (Indian Institute of Technology) JEE (Joint Entrance Exam). https://en.wikipedia.org/wiki/Joint_Entrance_Examination
52.
by d4rkp4ttern 4mo ago
Very very tangential, and at the risk of down-votes, the recent trend of X-articles (or whatever they call them) is extremely irksome. When I try to view on mobile it takes 3-4 hops to get to the article, and the articles always look hyper-
53.
by d4rkp4ttern 4mo ago
Is there anything similar to (or better than) overleaf for collaborating on typst docs? To be clear, overleaf isn’t all that great but I sometimes work with groups that are used to it. Have conferences that traditionally accepted latex sour
54.
by d4rkp4ttern 4mo ago
Yes CC prompt can be ~30K tokens. I definitely do not use this as a daily driver. I did use it a few times for sensitive document work with Qwen3.6 MOE.
55.
by d4rkp4ttern 4mo ago
It’s relatively simple to use llama.cpp/server to spin up a local LLM to work with Claude Code or Codex-CLI. The required llama server settings are often scattered all over so I maintain a set of instructions here for several popular o
56.
by d4rkp4ttern 4mo ago
Yes, and this is also why I haven’t yet tried the new “dynamic workflows” which spawn hundreds of agents that happily eat through your token limits.
57.
by d4rkp4ttern 4mo ago
Right, even a conservative 200k context length is on the order of 200 pages, which is more than enough context to arrive at an answer.
58.
by d4rkp4ttern 4mo ago
A related idea is to have the LLM quiz you, Socratic-style about a topic of interest. It persists in asking questions at deeper levels until you arrive at the answer yourself. This forces you to think hard about a problem, and this effort h
59.
by d4rkp4ttern 4mo ago
Yes, it can be used to run any script in another pane, so it can spawn Claude or codex for example. Also see the Tmux-cli “launch” sub-command , it will create a new pane and run a command.
60.
by d4rkp4ttern 4mo ago
This. I built a wrapper Tmux-cli to make this more reliable: https://pchalasani.github.io/claude-code-tools/tools/tmux-cl...

More ›