Hacker News: FranciscoCarlos

New comment by FranciscoCarlos in "Show HN: OpenTokenMini: Monitor and Fix Your AI Spending (FOSS)"

FranciscoCarlos — Mon, 19 Oct 2026 15:44:17 +0000

Author here, feel free to ask anything

Show HN: OpenTokenMini: Monitor and Fix Your AI Spending (FOSS)

FranciscoCarlos — Mon, 19 Oct 2026 15:44:01 +0000

Article URL: https://opentokenmini.org/

Comments URL: https://news.ycombinator.com/item?id=50153820

Points: 1

# Comments: 1

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Thu, 15 Oct 2026 12:45:49 +0000

Still a crutch around incomplete pre-training data, a ton of useful knowledge wont be properly represented on the weights.

3D Model Weights: Explore HuggingFace's top models in 3D

FranciscoCarlos — Thu, 15 Oct 2026 09:52:43 +0000

Article URL: https://franciscocarloserra.github.io/3d-model-weights/

Comments URL: https://news.ycombinator.com/item?id=50101410

Points: 2

# Comments: 0

New comment by FranciscoCarlos in "Cutoff Bench: How outdated is the training data of major LLMs"

FranciscoCarlos — Wed, 14 Oct 2026 16:48:25 +0000

Also a great potential indicator on when the pretraining run for the model started.

New comment by FranciscoCarlos in "Cutoff Bench: How outdated is the training data of major LLMs"

FranciscoCarlos — Wed, 14 Oct 2026 16:47:52 +0000

Willing to scale it up if there's interest

Cutoff Bench: How outdated is the training data of major LLMs

FranciscoCarlos — Wed, 14 Oct 2026 16:43:44 +0000

Article URL: https://franciscocarloserra.github.io/cutoff-bench/

Comments URL: https://news.ycombinator.com/item?id=50090642

Points: 3

# Comments: 3

New comment by FranciscoCarlos in "Muscriptor Large: Any Song to MIDI"

FranciscoCarlos — Wed, 14 Oct 2026 15:14:26 +0000

It works surprisingly well https://franciscocarloserra.github.io/mp3tomidi/

Ungated quant https://huggingface.co/DamRsn/muscriptor-gguf

Muscriptor Large: Any Song to MIDI

FranciscoCarlos — Wed, 14 Oct 2026 15:14:26 +0000

Article URL: https://huggingface.co/MuScriptor/muscriptor-large

Comments URL: https://news.ycombinator.com/item?id=50088901

Points: 4

# Comments: 1

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Wed, 14 Oct 2026 13:36:06 +0000

Yeah, it can be worked around with in-context instructions but they still tend to scope, delegate and architect without awareness of the fact that agents exist and work. It's an actual big deal IMHO, opus 5.5 is the most updated AFAIK.

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Wed, 14 Oct 2026 13:33:47 +0000

They know it and they try to cover it up with a fake reported cutoff date. Claude reports June '26 but it has only partial 2026 data (tested it against the API and it's 25 as well). GPT refuses to answer but it's old 25 data.

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Wed, 14 Oct 2026 08:20:21 +0000

Seems like a rushed RL run on top of an old model to jump into the hype train, seems reasonable since they hold a fantastic distribution moat. Even if it's not SOTA a billion people will use it and they'll collect data-points for future iterations. Same as muse.

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Wed, 14 Oct 2026 08:16:57 +0000

How is a model that has never seen an agent on it's pretraining data going to reasonably plan and fan-out tasks to multiple agents? Or engineer software for agents to edit semi-automatically, it has zero awareness of it's capabilities in it's internal representation.

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Tue, 13 Oct 2026 19:56:50 +0000

They need to pretrain on fresh data.

New comment by FranciscoCarlos in "Gemini 4 Argon"

FranciscoCarlos — Tue, 13 Oct 2026 19:28:18 +0000

Gemini 4 Argon

As of my knowledge cutoff, three of the top state-of-the-art large language models are Google's Gemini 2.5 Pro, OpenAI's o3 (and GPT-4.5), and Anthropic's Claude 3.7 Sonnet.