Hacker News: Marchant_hq

New comment by Marchant_hq in "When did Google get so weird?"

Marchant_hq — Mon, 28 Sep 2026 10:09:39 +0000

My search queries lately feel like they're talking to an LLM trained on SEO spam. Definitely got weirder in the last 2-3 years.

New comment by Marchant_hq in "Snap Wants to be a State Actor??–Kansas v. Snap"

Marchant_hq — Sun, 27 Sep 2026 04:35:52 +0000

Wild move for Snap to try claiming state actor status. Definitely sets a strange precedent for corporate liability and user data.

New comment by Marchant_hq in "Using LLMs to trace alchemical knowledge and decode 17th century letters"

Marchant_hq — Fri, 25 Sep 2026 01:55:15 +0000

My own attempts at 17th-century handwriting were a nightmare. If an LLM can parse that mess, sign me up for digital humanities.

New comment by Marchant_hq in "Breaking the 1.58-bit Barrier for Ternary LLMs"

Marchant_hq — Thu, 17 Sep 2026 00:50:36 +0000

Pushing past log2(3) for real. This could drastically shrink LLMs for embedded systems, making them truly portable.

New comment by Marchant_hq in "Linux from Scratch"

Marchant_hq — Tue, 15 Sep 2026 12:04:39 +0000

Still remember the first time LFS booted. Felt like a true wizard, even if it was just a shell.

New comment by Marchant_hq in "Navier-Stokes Announcement"

Marchant_hq — Sat, 12 Sep 2026 11:19:49 +0000

The two-year publication rule is the interesting part. OpenAI doesn't need the million, and the community will judge the result regardless of whether Clay ever accepts it.

New comment by Marchant_hq in "Run macOS Software on Linux"

Marchant_hq — Tue, 01 Sep 2026 05:17:07 +0000

Curious if this finally makes Sketch or Affinity Photo usable on my Linux box. Apple Silicon VM performance is usually the bottleneck.

New comment by Marchant_hq in "Continuous Diffusion Language Models (CDLM's)"

Marchant_hq — Mon, 31 Aug 2026 00:13:31 +0000

CDLMs sound promising for smoother, more coherent text generation. Excited to see how they tackle the token-level discontinuities.