Hacker News: danieltanfh95

New comment by danieltanfh95 in "Small models also found the vulnerabilities that Mythos found"

danieltanfh95 — Sun, 12 Apr 2026 03:22:11 +0000

No, it is evidence for that point. You could just rattle off every possible vulnerability and have the cheap model scan for it in the harness through a loop.

Note that I say cheap, not small, because small models may lack the reasoning needed, but some models are cheap enough but retain enough reasoning (ala Sonnet 3.7+)

New comment by danieltanfh95 in "Small models also found the vulnerabilities that Mythos found"

danieltanfh95 — Sun, 12 Apr 2026 02:14:19 +0000

People really lack imagination. The point here is that a dedicated attacker with a good harness and really cheap models can run the attack regardless. It's like portscan/url search attacks. They could run all of these against all codebases and clients. However, on the flip side, this also means we could run cheap models against every PR made, and do a thorough red-team security review.

None of these requires mythos. If anything we just need Opus 4.5+ that is not lobotomised.

The Case That More Openness Brings More Good to Society

danieltanfh95 — Sat, 11 Apr 2026 14:19:19 +0000

Article URL: https://danieltan.weblog.lol/2026/04/the-case-that-more-openness-brings-more-good-to-society

Comments URL: https://news.ycombinator.com/item?id=47730864

Points: 2

# Comments: 1

New comment by danieltanfh95 in "Show HN: Marimo pair – Reactive Python notebooks as environments for agents"

danieltanfh95 — Fri, 10 Apr 2026 16:22:12 +0000

interesting, i also went the same direction with https://github.com/danieltanfh95/replsh because i needed to support existing legacy python codebases.

New comment by danieltanfh95 in "Show HN: Marimo pair – Reactive Python notebooks as environments for agents"

danieltanfh95 — Fri, 10 Apr 2026 16:21:02 +0000

built https://github.com/danieltanfh95/replsh to pair with local python sessions without additional dependencies, allowing LLMs to directly ground their investigation and coding against local repos and environments. Now supporting docker as well, ssh support will come in the near future.

New comment by danieltanfh95 in "ML promises to be profoundly weird"

danieltanfh95 — Wed, 08 Apr 2026 16:44:55 +0000

I think the discussion has to be more nuanced than this. "LLMs still can't do X so it's an idiot" is a bad line of thought. LLMs with harnesses are clearly capable of engaging with logical problems that only need text. LLMs are not there yet with images, but we are improving with UI and access to tools like figma. LLMs are clearly unable to propose new, creative solutions for problems it has never seen before.

Give LLMs a Thinking Medium

danieltanfh95 — Wed, 08 Apr 2026 16:25:14 +0000

Article URL: https://github.com/danieltanfh95/replsh

Comments URL: https://news.ycombinator.com/item?id=47692400

Points: 1

# Comments: 0

Succession – Agent Lineage Evolution for an Agentic World

danieltanfh95 — Tue, 07 Apr 2026 00:55:44 +0000

Article URL: https://danieltan.weblog.lol/2026/04/succession-ale-for-an-agentic-world

Comments URL: https://news.ycombinator.com/item?id=47669404

Points: 1

# Comments: 0

New comment by danieltanfh95 in "Signals, the push-pull based algorithm"

danieltanfh95 — Mon, 06 Apr 2026 07:25:02 +0000

How is error handling expected to happen here?

The Armor Gap

danieltanfh95 — Tue, 24 Mar 2026 10:55:27 +0000

Article URL: https://danieltan.weblog.lol/2026/03/the-armor-gap

Comments URL: https://news.ycombinator.com/item?id=47500862

Points: 1

# Comments: 0

Pokemon and the Iran War Nerdsniped Me into Quantifying Strategy

danieltanfh95 — Sat, 21 Mar 2026 15:24:50 +0000

Article URL: https://danieltan.weblog.lol/2026/03/how-pokemon-and-the-iran-war-nerdsniped-me-into-quantifying-strategy

Comments URL: https://news.ycombinator.com/item?id=47467870

Points: 2

# Comments: 0

New comment by danieltanfh95 in "We Have Learned Nothing"

danieltanfh95 — Thu, 19 Mar 2026 08:46:47 +0000

Maybe the more rational conclusion is that it is purely a random chance that a startup would succeed, so we should just increase the amount of startups in the first place instead of restricting it.

New comment by danieltanfh95 in "Nasdaq's Shame"

danieltanfh95 — Mon, 16 Mar 2026 02:22:21 +0000

Really the same mechanics with crypto

New comment by danieltanfh95 in "What is agentic engineering?"

danieltanfh95 — Mon, 16 Mar 2026 02:20:09 +0000

Agentic engineering is working from documentation -> code and automating the translation process via agents. This is distinct from the waterfall process which describes the program, but not the code itself, and waterfall documentation cannot be translated directly to code. Agent plans and session have way more context and details that are not captured in waterfall due to differences in scope.

New comment by danieltanfh95 in "Step 3.5 Flash – Open-source foundation model, supports deep reasoning at speed"

danieltanfh95 — Tue, 24 Feb 2026 04:10:13 +0000

It is just search. If it can't even execute search then what else would it be good at?

New comment by danieltanfh95 in "Step 3.5 Flash – Open-source foundation model, supports deep reasoning at speed"

danieltanfh95 — Thu, 19 Feb 2026 05:05:38 +0000

Hallucinates like crazy. use with caution. Tested it with a simple "Find me championship decks for X pokemon", "How does Y deck work". Opus 4.6, Deepseek and Kimi all performed well as expected.

New comment by danieltanfh95 in "Parse, Don't Validate (2019)"

danieltanfh95 — Wed, 11 Feb 2026 02:45:17 +0000

> static types are not about “classifying the world” or pinning down the structure of every value in a system. The reality is that static type systems allow specifying exactly how much a component needs to know about the structure of its inputs, and conversely, how much it doesn’t.

New comment by danieltanfh95 in "Parse, Don't Validate (2019)"

danieltanfh95 — Wed, 11 Feb 2026 02:17:31 +0000

python, js, ruby, clojure, erlang etc are created by senior engineers who are fully aware that haskell etc exists, so have you considered that people you claim are "senior" aren't senior enough?

New comment by danieltanfh95 in "Parse, Don't Validate (2019)"

danieltanfh95 — Wed, 11 Feb 2026 02:09:25 +0000

People keep saying this and yet in the decades of my career the industry bounces between being fully dynamic and fully typed according to the affordability of senior engineers. What you are saying are covered by tests, not types.

New comment by danieltanfh95 in "Parse, Don't Validate (2019)"

danieltanfh95 — Wed, 11 Feb 2026 02:06:30 +0000

"good type systems" in Python and Javascript has a escape hatch via dicts/objects, and are frankly, not "good" compared to FP via haskell etc, but it is realistic, reliable, open and consistent, which is exactly my point!

You could spend more time understanding Pydantic before you make this comment however.