Hacker News: bbcc90

New comment by bbcc90 in "How We Broke Top AI Agent Benchmarks: And What Comes Next"

bbcc90 — Sat, 11 Apr 2026 21:07:03 +0000

Yes good evals are really hard - that’s not really news.

This team is doing a good job. They use problems that were created in last 30days to avoid training set leakage. https://swe-rebench.com/

New comment by bbcc90 in "430k-year-old well-preserved wooden tools are the oldest ever found"

bbcc90 — Wed, 28 Jan 2026 14:55:15 +0000

Sure, but did they have MCP?

New comment by bbcc90 in "Gemma 3 270M: Compact model for hyper-efficient AI"

bbcc90 — Fri, 15 Aug 2025 05:24:37 +0000

it does work; just download from HF and load in the app

New comment by bbcc90 in "Show HN: Immersive Gaussian Splat experience of Sutro Tower, San Francisco"

bbcc90 — Fri, 21 Feb 2025 03:23:06 +0000

true. See here for a map: https://sfplanninggis.s3.amazonaws.com/hub/BIGmap.pdf

I live in the that space (between tower and the city) and the local neighbourhood group (HANC) is ridiculously NIMBY. rezoning is happening but it's slow going...

New comment by bbcc90 in "Tensor Product Attention Is All You Need"

bbcc90 — Wed, 22 Jan 2025 07:12:58 +0000

(trying to move the critique beyond the title...)

When trying to deploy llms in with larger context windows constrained environments 2 things start to hurt: a) increased memory footprint for longer KV cache b) increased decode speed due to longer context window. this paper addresses a) only, which is useful, but we are still left with b) (right?)

New comment by bbcc90 in "Microsoft acquires twice as many Nvidia AI chips as tech rivals"

bbcc90 — Wed, 18 Dec 2024 17:48:02 +0000

This is a bug not a feature:

https://finance.yahoo.com/news/microsoft-stock-receives-rare...

New comment by bbcc90 in "New Nvidia driver adds 16bit float support to OpenCL [pdf]"

bbcc90 — Sat, 17 Feb 2024 06:44:19 +0000

any idea why this took them so long?

New comment by bbcc90 in "Sam Altman Seeks Trillions of Dollars to Reshape Business of Chips and AI"

bbcc90 — Fri, 09 Feb 2024 08:48:02 +0000

If you are going to go vertical then do it properly.

OpenAI could just build their own framework for internal use that works well on their silicon (see Jax+tpu)

Their starting point? Triton plus some triton libs. Jax chipped away at TF like this, and no reason why Triton can’t do the same to PyTorch.

New comment by bbcc90 in "How to enjoy SF"

bbcc90 — Sun, 06 Aug 2023 23:34:38 +0000

Love this. I’m 10 years in SF myself. One thing I would add is the people

New comment by bbcc90 in "Growing from engineer to manager"

bbcc90 — Sun, 11 Jun 2023 16:05:24 +0000

I would also say that being an engineer at L8 is a meaningfully different set of skills than L7. The whole ‘what got you here won’t get you there idea’ starts at L7 for either track…

New comment by bbcc90 in "Mini-Consortia Forming Around Chiplets"

bbcc90 — Fri, 31 Mar 2023 06:29:42 +0000

Chiplets and standardized hardware interfaces are a great start.

Now we just need to solve the remaining 90%: firmware, software, security, virtualization, …, …