Hacker News: hrpnk

New comment by hrpnk in "Strands Decider 2B: a small, open-source, decision model"

hrpnk — Wed, 07 Oct 2026 06:21:20 +0000

clef from cloudflare runs on llama.cpp - being locked-in to strands cli would be a bummer and will slow down adoption.

Since it's a LoRa on Qwen, I assume this is runnable via llama.cpp. Pity that the PEFT/LoRa->GGUF translation is left to the user. Anyone got past:

    $ uv run --with transformers==5.19.0 convert_lora_to_gguf.py ~/Downloads/lora --dry-run --verbose
    [...]
      File "/Users/user/repos/llama.cpp/conversion/base.py", line 630, in map_tensor_name
    raise ValueError(f"Can not map tensor {name!r}")
    ValueError: Can not map tensor 'layers.0.linear_attn.in_proj_a.weight'

New comment by hrpnk in "GLM-5.3 and the spread of advanced cyber capabilities"

hrpnk — Tue, 29 Sep 2026 20:36:59 +0000

Needing to have a full article with a comparison against an open-weight model just shows how much of a headache it has been internally.

New comment by hrpnk in "Revolut confirms customer data breach through fake government requests"

hrpnk — Sun, 13 Sep 2026 18:17:06 +0000

If you are forced to ID check with the bank via a 3rd party, the only way is to ask for removal of data after the ID check. Do you see other ways?

New comment by hrpnk in "David Sacks: OpenAI and Anthropic Don't Need Regulations to Pace Frontier Models"

hrpnk — Sun, 13 Sep 2026 18:15:40 +0000

I wished OpenAI and Anthropic IPOd already. They just try to outcompete each other by finding the most splashy headlines undercutting prices in unhealthy and unsustainable means long-term.

New comment by hrpnk in "Revolut confirms customer data breach through fake government requests"

hrpnk — Sun, 13 Sep 2026 11:19:39 +0000

Even if the trigger was spoofed, how come there is no secure channel that the govt provides to receive the data? Was this one also compromised?

Commerce Agents: Reference blueprint for building shopping and merchant agents

hrpnk — Wed, 02 Sep 2026 19:56:29 +0000

Article URL: https://github.com/anthropics/commerce-agents

Comments URL: https://news.ycombinator.com/item?id=49541574

Points: 3

# Comments: 0

New comment by hrpnk in "An ongoing 3D-printer AGPL violation"

hrpnk — Wed, 26 Aug 2026 19:34:40 +0000

What would the company need to do to comply with licensing? Release the studio, shared library, and server-side components as AGPL?

Agentic Engineering at Zalando: A Snapshot

hrpnk — Tue, 18 Aug 2026 18:15:49 +0000

Article URL: https://engineering.zalando.com/posts/2026/08/agentic-engineering-at-zalando-a-snapshot.html

Comments URL: https://news.ycombinator.com/item?id=49350045

Points: 3

# Comments: 0

New comment by hrpnk in "Everyone is building LLM routers, we deprecated ours"

hrpnk — Fri, 31 Jul 2026 20:38:47 +0000

Even the simplest router that differentiates models between plan & execute will ensure that this is consistently followed. Folks have too high FOMO to choose themselves.

New comment by hrpnk in "Claude Is Down"

hrpnk — Wed, 29 Jul 2026 20:35:02 +0000

- write/update tickets & collect prompt queues for when it's back up

- read & respond to customer feedback

New comment by hrpnk in "Claude Opus 5"

hrpnk — Fri, 24 Jul 2026 19:23:38 +0000

The breaking changes vs. Opus 4.8 are interesting [1]

1. Thinking on by default: On Claude Opus 4.8, requests without a thinking field run without thinking; on Claude Opus 5, the same requests run with adaptive thinking.

2. Disabling thinking is capped at high effort: You can still turn thinking off with thinking: {type: "disabled"}, but only at an effort level of high or below.

[1] https://platform.claude.com/docs/en/about-claude/models/migr...

New comment by hrpnk in "Please don't discontinue Gemini 2.5 Flash"

hrpnk — Fri, 10 Jul 2026 23:18:50 +0000

The smell is about folks being attached to the same model because they're used to it. Just like any UI change in an app that one uses everyday is annoying.

What's often missing is the actual engineering. If you have evals for your use case, you use these to adjust to the new model. DSPy has great prompt engineering constructs it ships with.

The smell is all about vibing, where a model feels better because the structure of the answers is more familiar to a person, instead of engineering where given constraints and input/outputs one is using/bending the system to fulfil the requirements.

New comment by hrpnk in "Please don't discontinue Gemini 2.5 Flash"

hrpnk — Fri, 10 Jul 2026 21:07:00 +0000

I love how there is a "Please do not discontinue gemini-2.0-flash[-lite], 2.5 is NOT an equivalent" from Feb 20th. Getting too attached to models is a smell.

New comment by hrpnk in "GPT-5.6"

hrpnk — Thu, 09 Jul 2026 19:27:18 +0000

They highlight the cache write price now much more in the guide. Did it increase vs. prior generations?

New comment by hrpnk in "Kimi K2.7 Code is generally available in GitHub Copilot"

hrpnk — Thu, 02 Jul 2026 19:42:19 +0000

There is literally no more reason to use their service. I don't need a monthly expiring token pool drained at API pricing which I can get from any provider.

New comment by hrpnk in "Kimi K2.7 Code is generally available in GitHub Copilot"

hrpnk — Thu, 02 Jul 2026 19:40:07 +0000

Wonder if this is a product decision or rather a limitation on GPUs. Their Inf2 instances require AWS Neuron [1]. Maybe porting models to run on their hardware takes too much time.

[1] https://aws.amazon.com/ai/machine-learning/neuron/

New comment by hrpnk in "Open Code Review – An AI-powered code review CLI tool"

hrpnk — Fri, 05 Jun 2026 07:36:51 +0000

Rule files are in https://github.com/alibaba/open-code-review/tree/main/intern... (in Chinese)

Adrianco's Retort: measure how reliable, fast and expensive your LLM is

hrpnk — Thu, 04 Jun 2026 19:10:55 +0000

Article URL: https://adrianco.medium.com/how-reliable-fast-and-expensive-is-each-version-of-claude-code-sonnet-through-opus-4-8-fast-272d74ffc869

Comments URL: https://news.ycombinator.com/item?id=48403232

Points: 2

# Comments: 0

New comment by hrpnk in "Uber's $1,500/month AI limit is a useful signal for AI tool pricing"

hrpnk — Wed, 03 Jun 2026 18:54:49 +0000

If budgeted at $1,500/month per user, power users still can get 5-10x of that allocation if the user pool is large enough.

New comment by hrpnk in "Uber's $1,500/month AI limit is a useful signal for AI tool pricing"

hrpnk — Wed, 03 Jun 2026 18:53:30 +0000

$1500/mth is token pricing.

Your other plans are fixed price with rate limits where you get more tokens than the dollar equivalent you pay monthly. These plans are economical only if majority of users spend less tokens in $ than the plan's costs. This subsidizes the gap vs. power users who spend multiple k$ monthly in API tokens.