Hacker News: kp1197

I turned Jev into a (lousy) chatbot

kp1197 — Sun, 20 Sep 2026 17:51:27 +0000

Article URL: https://github.com/kyle-pena-nlp/jevchat/

Comments URL: https://news.ycombinator.com/item?id=49778162

Points: 77

# Comments: 26

P(Kill-Switch|Detection)

kp1197 — Mon, 14 Sep 2026 18:55:22 +0000

Article URL: https://www.lesswrong.com/posts/tvaQniyER4BmsQpbq/p-kill-switch-or-detection

Comments URL: https://news.ycombinator.com/item?id=49702077

Points: 1

# Comments: 0

P(Kill-Switch|Detection)

kp1197 — Mon, 31 Aug 2026 13:18:34 +0000

Article URL: https://www.lesswrong.com/posts/tvaQniyER4BmsQpbq/p-kill-switch-or-detection

Comments URL: https://news.ycombinator.com/item?id=49509433

Points: 2

# Comments: 0

Prompt Sufficiency: A Missive for the Managerial Class

kp1197 — Sun, 23 Aug 2026 12:30:58 +0000

Article URL: https://www.lesswrong.com/posts/jDHc4oK4oPwfh6TcK/prompt-sufficiency-a-missive-for-the-managerial-class

Comments URL: https://news.ycombinator.com/item?id=49408313

Points: 1

# Comments: 0

New comment by kp1197 in "The Napoleon Technique: Postponing things to increase productivity"

kp1197 — Sat, 10 Jan 2026 07:47:52 +0000

Hello! Have you ever felt like too important to answer questions? Guess what! This one weird trick will turn you into a French General in the boardroom!

New comment by kp1197 in "Straussian Memes"

kp1197 — Sun, 04 Jan 2026 16:06:36 +0000

I kind of regret calling it Straussian because of the baggage that comes with that label. The key differentiator here compared to exoteric/esoteric is the stabilization mechanism: social costs to either upgrading one’s own understanding, or upgrading someone else’s understanding. That’s what keeps the readings differentiated and therefore the tiered structure stable. Keeping the structure encourages the memes survival, because it widens the acceptability of the message to meet audiences where they are. I think I need a few more examples and I ought to make this point a little more explicit.

Straussian Memes

kp1197 — Thu, 01 Jan 2026 21:24:39 +0000

Article URL: https://www.lesswrong.com/posts/CAwnnKoFdcQucq4hG/straussian-memes-a-lens-on-techniques-for-mass-persuasion

Comments URL: https://news.ycombinator.com/item?id=46458165

Points: 43

# Comments: 52

New comment by kp1197 in "Using TypeScript to obtain one of the rarest license plates"

kp1197 — Fri, 19 Dec 2025 15:27:43 +0000

Soon he may be making vanity license plates

Toddler Shoggoth Has Plenty of Raw Material (The Memetic Cocoon Threat Model)

kp1197 — Mon, 17 Nov 2025 17:17:40 +0000

Article URL: https://www.lesswrong.com/posts/zvkjQen773DyqExJ8/toddler-shoggoth-has-plenty-of-raw-material-the-memetic

Comments URL: https://news.ycombinator.com/item?id=45955680

Points: 2

# Comments: 1

New comment by kp1197 in "Unconventional Ways to Cast in TypeScript"

kp1197 — Fri, 24 Oct 2025 18:35:39 +0000

Isn't the post more about the interesting capabilities of the language rather than an engineering context?

New comment by kp1197 in "React is winning by default and slowing innovation"

kp1197 — Tue, 16 Sep 2025 04:25:32 +0000

I remember the days of countless insane, boutique javascript libraries. React's ubiquity is a victory - let's not encourage throwing something away in the name of some vague notion of "innovation".

New comment by kp1197 in "Gradient Descent on Token Input Embeddings"

kp1197 — Fri, 18 Jul 2025 18:15:53 +0000

Does performing gradient descent on token input embeddings lead to interpretable results? And if not, why?

Gradient Descent on Token Input Embeddings

kp1197 — Fri, 18 Jul 2025 18:15:53 +0000

Article URL: https://www.lesswrong.com/posts/GK2LSzxjEejzDjzDs/gradient-descent-on-token-input-embeddings

Comments URL: https://news.ycombinator.com/item?id=44608055

Points: 3

# Comments: 1

New comment by kp1197 in "Compiling LLMs into a MegaKernel: A path to low-latency inference"

kp1197 — Thu, 19 Jun 2025 21:06:09 +0000

After working pretty closely with vLLM and SGLang over the past few months, this is EXACTLY what I had envisioned what a successor project would look like - analyzing an operation dependency graph and then fusing (or, at a minimum, scheduling tasks smarter). Congrats to the team.

New comment by kp1197 in "Me an' Algernon – grappling with (temporary) cognitive decline"

kp1197 — Sun, 15 Jun 2025 20:50:02 +0000

It is unclear to me is whether Kent recovered his cognitive ability or not. He says he has, "thanks to the genie" (which seems to be his name for coding LLMs). Does this mean that with a genie, he is now able to produce like he was before? Or that he used a genie to help "re-train" himself?

New comment by kp1197 in "V-JEPA 2 world model and new benchmarks for physical reasoning"

kp1197 — Wed, 11 Jun 2025 20:40:24 +0000

Robots that can do anything.

New comment by kp1197 in "Elon Musk joined Trump-Zelensky call amid concerns for future of Ukraine war"

kp1197 — Sat, 09 Nov 2024 04:10:33 +0000

Is it though? He's perfectly positioned to do so.

New comment by kp1197 in "Blockchain-Backed Click App Reinvents Trust in Photojournalism"

kp1197 — Wed, 06 Nov 2024 21:44:59 +0000

As I've posted elsewhere in the comments here, I think there are good solutions, but it will take some cooperation to get them off the ground (adoption of standards, licenses, etc.). Here is what I've proposed, which comes pretty close to what the CAI (Content Authenticity Initiative) is doing: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...

New comment by kp1197 in "Blockchain-Backed Click App Reinvents Trust in Photojournalism"

kp1197 — Wed, 06 Nov 2024 19:05:51 +0000

I've got some good solutions on this front, I think: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...

New comment by kp1197 in "DeepMind debuts watermarks for AI-generated text"

kp1197 — Wed, 06 Nov 2024 18:35:16 +0000

Watermarking is not the way to go. It relies on the honesty of the producers, and watermarks can be easily stripped. With images, the way to go is detect authentic images, not fake ones. I've written about this extensively: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...