Hacker News: joshred

New comment by joshred in "Uber torches 2026 AI budget on Claude Code in four months"

joshred — Fri, 01 May 2026 18:55:10 +0000

If you think about it, the production quality is probably log-linear, so the token growth may well be exponential.

New comment by joshred in "Google releases Gemma 4 open models"

joshred — Thu, 02 Apr 2026 18:08:07 +0000

From what I've read, that's already part of their training. They are scored based on each step of their reasoning and not just their solution. I don't know if it's still the case, but for the early reasoning models, the "reasoning" output was more of a GUI feature to entertain the user than an actual explanation of the steps being followed.

New comment by joshred in "Telnyx package compromised on PyPI"

joshred — Fri, 27 Mar 2026 19:48:26 +0000

This is complete nonsense. pip has all the same problems that you say uv has.

New comment by joshred in "Local AI is driving the biggest change in laptops in decades"

joshred — Tue, 23 Dec 2025 21:33:01 +0000

Paying $30-$70/day to commute is economical?

New comment by joshred in "Australia begins enforcing world-first teen social media ban"

joshred — Wed, 10 Dec 2025 20:52:03 +0000

Gerrymandering already exists. Voter suppression was huge in the past, and may become huge again. The supreme court made sure of that.

And also... the supreme court keeps issuing partisan decisions.

So... what is left? Number 3?

I guess you're arguing that federalism protects people, but how does it do that in a way that isn't already being eroded?

New comment by joshred in "Republican push to make U.S. census surveys voluntary alarms statisticians"

joshred — Mon, 17 Nov 2025 18:34:07 +0000

I work for state government. We've used the ACS survey to try and determine whether we were unfairly targeting non-native English speakers with some of our decisions. It's also used a lot in academia.

If I had to guess, commercial organizations have access to more invasive and higher quality data that they obtain through credit card companies, lexus-nexus or other data brokers. This attitude mostly harms organizations involved in the social sciences.

New comment by joshred in "F5 says hackers stole undisclosed BIG-IP flaws, source code"

joshred — Wed, 15 Oct 2025 18:51:25 +0000

BRB, changing handle to 'nation-state'. Need the resume fodder.

New comment by joshred in "Le Chat: Custom MCP Connectors, Memories"

joshred — Thu, 18 Sep 2025 19:17:41 +0000

How do you get the alternate tokens?

New comment by joshred in "Le Chat: Custom MCP Connectors, Memories"

joshred — Thu, 04 Sep 2025 18:04:03 +0000

It sounds like they are describing a regex filter being applied to the model's beam search. LLMs generate the most probable words, but they are frequently tracking several candidate phrases at a time and revising their combined probability. It lets them self correct if a high probability word leads to a low probability phrase.

I think they are saying that if highest probability phrase fails the regex, the LLM is able to substitute the next most likely candidate.

New comment by joshred in "Show HN: Whispering – Open-source, local-first dictation you can trust"

joshred — Mon, 18 Aug 2025 18:57:06 +0000

They might not now how whisper works. I suspect that the answer to their question is 'yes' and the reason they can't find a straightforward answer through your project is that the answer is so obvious to you that it's hardly worth documenting.

Whisper for transcription tries to transform audio data into LLM output. The transcripts generally have proper casing, punctuation and can usually stick to a specific domain based on the surrounding context.

New comment by joshred in "TREAD: Token Routing for Efficient Architecture-Agnostic Diffusion Training"

joshred — Mon, 18 Aug 2025 18:49:14 +0000

This is the high-level explanation of the simplest diffusion architecture. The model trains by taking an image and iteratively adding noise to the image until there is only noise. Then they take that sequence of noisier and noisier images and they reverse it. The result is that they start with only noise, and they predict the removal of noise at step until they get to the final step (which should be the original image (or training input)).

That process means they may require a hundred or more training iterations on a single image. I haven't digested the paper, but it sounds like they are proposing something conceptually similar to skip layers (but significantly more involved).

New comment by joshred in "GitHub was having issues"

joshred — Tue, 12 Aug 2025 15:46:43 +0000

Sooo....

What's the consensus on gitlab?

New comment by joshred in "Claude Code is all you need"

joshred — Mon, 11 Aug 2025 17:59:01 +0000

I think they're fantastic at generating the sort of thing I don't like writing out. For example, a dictionary mapping state names to their abbreviations, or extracting a data dictionary from a pdf so that I can include it with my documentation.

New comment by joshred in "Why “Move Fast and Break Things” Doesn’t Work Anymore"

joshred — Sat, 04 Jan 2020 15:25:51 +0000

Hey quibbler, you defend a castle (not a moat).

New comment by joshred in "Sourcetrail, interactive source explorer, is now free and open-source"

joshred — Tue, 19 Nov 2019 12:18:38 +0000

If you say so...

New comment by joshred in "Court: Suspicionless Searches of Travelers’ Phones and Laptops Unconstitutional"

joshred — Tue, 12 Nov 2019 22:39:17 +0000

It's already pretty well established as law. There's not really any room to argue.

New comment by joshred in "Twitter to ban political advertising"

joshred — Wed, 30 Oct 2019 23:48:48 +0000

Isn't organic reach a little more meritocratic than just throwing money into paid ads?

New comment by joshred in "Larry Wall has approved renaming Perl 6 to Raku"

joshred — Fri, 11 Oct 2019 21:04:55 +0000

A camel strikes me as unsophisticated yet perfect.

New comment by joshred in "FBI Says QAnon, Internet Conspiracy Theorists Are National Security Threats"

joshred — Fri, 02 Aug 2019 00:09:31 +0000

That's a bad argument.

If I were to call you a moron, that doesn't mean I'm advocating removing you from the gene pool.

New comment by joshred in "Firefox to add Tor Browser anti-fingerprinting technique called letterboxing"

joshred — Wed, 06 Mar 2019 23:52:27 +0000

Does that matter? Don't devs cater to the outliers?