Hacker News: alex000kim

Silverback Imfura took a chance, and ended up alone

alex000kim — Sun, 10 May 2026 02:06:19 +0000

Article URL: https://gorillafund.org/mountain-gorillas/silverback-imfura-took-a-chance-and-ended-up-alone/

Comments URL: https://news.ycombinator.com/item?id=48080316

Points: 75

# Comments: 24

Do AI Detectors Work Well Enough to Trust?

alex000kim — Mon, 04 May 2026 00:10:36 +0000

Article URL: https://www.chicagobooth.edu/review/do-ai-detectors-work-well-enough-trust

Comments URL: https://news.ycombinator.com/item?id=48003026

Points: 5

# Comments: 1

You've been doing harness engineering all along

alex000kim — Sun, 26 Apr 2026 18:13:46 +0000

Article URL: https://alex000kim.com/posts/2026-04-26-harness-engineering/

Comments URL: https://news.ycombinator.com/item?id=47912495

Points: 7

# Comments: 0

Lessons from Going Solo

alex000kim — Mon, 20 Apr 2026 14:32:13 +0000

Article URL: https://alex000kim.com/posts/2026-04-20-lessons-from-going-solo/

Comments URL: https://news.ycombinator.com/item?id=47834955

Points: 7

# Comments: 1

New comment by alex000kim in "Research-Driven Agents: When an agent reads before it codes"

alex000kim — Thu, 09 Apr 2026 19:46:02 +0000

That directory is huge already! I guess the index.md helps the agent find what it needs, but even the markdown file is very long - this would consume a ton of tokens.

Also I wonder who/what decides what papers go in there.

In the blog post, the agent is allowed to do its own search.

New comment by alex000kim in "Research-Driven Agents: What Happens When Your Agent Reads Before It Codes"

alex000kim — Thu, 09 Apr 2026 19:32:08 +0000

yup, as the blog says

> The full setup works with any project that has a benchmark and test suite.

so having a clear and measurable verification step is key. Meaning you can't simply give an AI agent a vague goal e.g. "improve the quality of the codebase" because it's too general.

New comment by alex000kim in "Research-Driven Agents: What Happens When Your Agent Reads Before It Codes"

alex000kim — Thu, 09 Apr 2026 19:28:53 +0000

I am sure this would works well in general. There is a challenge wrt to how to make them communicate effectively to e.g. 1) avoid duplicative work and 2) allow them to combine/overlay each others' findings to yield even better results

New comment by alex000kim in "Research-Driven Agents: When an agent reads before it codes"

alex000kim — Thu, 09 Apr 2026 19:26:18 +0000

sounds similar to "LLM Knowledge Bases" https://xcancel.com/karpathy/status/2039805659525644595

New comment by alex000kim in "The Claude Code Source Leak: fake tools, frustration regexes, undercover mode"

alex000kim — Tue, 31 Mar 2026 18:39:51 +0000

technically you're correct, but look at the prompt https://github.com/alex000kim/claude-code/blob/main/src/util...

it's written to _actively_ avoid any signs of AI generated code when "in a PUBLIC/OPEN-SOURCE repository".

Also, it's not about you. Undercover mode only activates for Anthropic employees (it's gated on USER_TYPE === 'ant', which is a build-time flag baked into internal builds).

New comment by alex000kim in "The Claude Code Source Leak: fake tools, frustration regexes, undercover mode"

alex000kim — Tue, 31 Mar 2026 18:11:12 +0000

Oh right, I just saw https://news.ycombinator.com/item?id=47582220 will update the post with this link

The Claude Code Source Leak: fake tools, frustration regexes, undercover mode

alex000kim — Tue, 31 Mar 2026 13:04:30 +0000

Related ongoing thread: Claude Code's source code has been leaked via a map file in their NPM registry - https://news.ycombinator.com/item?id=47584540

Also related: https://www.ccleaks.com

Comments URL: https://news.ycombinator.com/item?id=47586778

Points: 1376

# Comments: 578

New comment by alex000kim in "Migrating from Slurm to Kubernetes"

alex000kim — Wed, 11 Feb 2026 18:51:27 +0000

Author here. I've seen the docs you linked to: Slurm uses "gang scheduling" to mean something specific (timesliced oversubscription where jobs alternate on shared resources).

I'm using the term in its broader CS sense: all-or-nothing co-scheduling of related processes across multiple processors [1]. This is the definition used across the K8s ecosystem e.g. Volcano [2], Kueue [3], and its Coscheduling plugin all define gang scheduling as "all or nothing" allocation.

I still stand by the origianl claim:

Slurm allocates multi-node jobs atomically, while vanilla K8s doesn't. its default scheduler places pods as resources become available, leading to partial allocations and deadlocks for distributed training. It's just a terminology clash. Thanks for the comment anyway.

[1] https://en.wikipedia.org/wiki/Gang_scheduling [2] https://volcano.sh/en/docs/plugins/ [3] https://www.coreweave.com/blog/kueue-a-kubernetes-native-sys...