Hacker News: k__

New comment by k__ in "DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis"

k__ — Fri, 31 Jul 2026 20:37:46 +0000

Haha, and I almost felt bad after seeing this chart yesterday.

New comment by k__ in "DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis"

k__ — Fri, 31 Jul 2026 14:10:23 +0000

Half OT:

Why do the cache hit rates seem to vary so much between harnesses?

I use pi, which is very minimalist, and I get a hit rate of ~99%. Paying like $1 a day for Flash. Yet, the hit rate mentioned on OpenRouter is only ~79%.

New comment by k__ in "DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis"

k__ — Fri, 31 Jul 2026 11:35:13 +0000

On OpenRouter it's 93 TPS.

New comment by k__ in "DeepSeek-V4-Flash Update"

k__ — Fri, 31 Jul 2026 11:11:50 +0000

I'm using pi and my caching is ~99%.

New comment by k__ in "DeepSeek-V4-Flash Update"

k__ — Fri, 31 Jul 2026 10:21:54 +0000

Yeah, it needs quite some hand holding.

I didn't do much agent coding and had a mix experience.

1. It would build something that was in the spirit of what I wanted, but unusable in practice.

2. It would build something quite useful, but only the public APIs were nice, the deeper code layers would get more and more convoluted.

3. It would built what I wanted and it would have okay-ish code.

However, for 3. I also had to add a custom AGENTS.md, many more code example, extra repos as subtrees, and review any code that had new concepts.

Much more work, but still much less than typing it all by hand.

New comment by k__ in "DeepSeek-V4-Flash Update"

k__ — Fri, 31 Jul 2026 10:03:49 +0000

I'd take more throughput while everything else stays the same.

New comment by k__ in "Kimi K3-256k"

k__ — Thu, 30 Jul 2026 08:14:06 +0000

While I tend to clear my session after every task, I feel less stressed when my context is as 10% than when it's at 30%

New comment by k__ in "Kimi K3-256k"

k__ — Thu, 30 Jul 2026 08:12:27 +0000

They remove potentially irrelevant details.

New comment by k__ in "Kimi K3-256k"

k__ — Thu, 30 Jul 2026 08:07:02 +0000

As I understand it, they would have to train a whole knew model to hard cap it's context to different lengths. That would be cheaper to train and had cheaper inf, but still a huge investment.

So I'd guess it's API level.

New comment by k__ in "Kimi-K3 Releases on HuggingFace 7/27"

k__ — Mon, 27 Jul 2026 12:44:30 +0000

Right

New comment by k__ in "The front end framework for correctness: built on Effect, architected like Elm"

k__ — Sat, 25 Jul 2026 17:33:28 +0000

Yeah, I think the Effect team tried building a compiler for once (TS++ or something) but they abandoned it, as it was too much work.

New comment by k__ in "Writing by hand is good for your brain"

k__ — Thu, 23 Jul 2026 16:29:07 +0000

Do crosswords count?

New comment by k__ in "Qwen 3.8"

k__ — Sun, 19 Jul 2026 15:01:34 +0000

My 2 weeks with DeepSeek V4:

Pro is ~50% more expensive than Flash.

Both need babysitting.

Plan, split in small tasks, give it docs, types, tests, linter, best practice examples, etc.

Always start a new session when starting a task.

Do regular manual sanity checks, and tell it to find issues in the codebase.

I pay like $1,50 per day for Pro.

New comment by k__ in "The Kimi K3 Moment"

k__ — Sat, 18 Jul 2026 19:24:10 +0000

Thanks!

What is the parento frontier?

New comment by k__ in "The Kimi K3 Moment"

k__ — Sat, 18 Jul 2026 18:50:04 +0000

Half-OT: can anyone recommend a LLM cost calculator that's up to date?

New comment by k__ in "Abstracting Effects with Continuations"

k__ — Fri, 17 Jul 2026 09:23:36 +0000

Effect's use of generators is also nice. While looking a bit off, it allows writing imperative business logic.

New comment by k__ in "Pebble Mega Update – July 2026"

k__ — Fri, 17 Jul 2026 08:45:13 +0000

Are they still using eink?

New comment by k__ in "German AI consortium releases Soofi S, an open 30B model that tops benchmarks"

k__ — Thu, 16 Jul 2026 19:12:37 +0000

"Fully open source model" with "Long-term, license-free availability for industry"

Nice trick to be not comparable with most other LLMs on the market, open weight or proprietary.

But I think, that's the right way.

New comment by k__ in "Inkling: Our Open-Weights Model"

k__ — Thu, 16 Jul 2026 00:04:38 +0000

I tried Hy3 today and liked it. It's a (small) step up from DSV4P.

Something on that level but multi-modal would be quite nice!

New comment by k__ in "Punch yourself in the face with reality"

k__ — Tue, 14 Jul 2026 20:17:30 +0000

While it's true that unsupervised AIs build a heap of trash, I have to admit, it doesn't look much worse than some of the code I saw powering products that sold for millions.

The only thing that changed is the scale.