Hacker News: timbilthttps://news.ycombinator.com/user?id=timbiltHacker News RSShttps://hnrss.org/hnrss v2.1.1Tue, 07 Apr 2026 23:59:00 +0000<![CDATA[NanoClaw's architecture is a masterclass in doing less]]>Article URL: https://jonno.nz/posts/nanoclaw-architecture-masterclass-in-doing-less/

Comments URL: https://news.ycombinator.com/item?id=47675254

Points: 44

# Comments: 11

]]>Tue, 07 Apr 2026 13:41:50 +0000https://jonno.nz/posts/nanoclaw-architecture-masterclass-in-doing-less/timbilthttps://news.ycombinator.com/item?id=47675254https://news.ycombinator.com/item?id=47675254<![CDATA[Google open-sources experimental agent orchestration testbed Scion]]>https://googlecloudplatform.github.io/scion/overview/

Comments URL: https://news.ycombinator.com/item?id=47675213

Points: 147

# Comments: 41

]]>Tue, 07 Apr 2026 13:39:15 +0000https://www.infoq.com/news/2026/04/google-agent-testbed-scion/timbilthttps://news.ycombinator.com/item?id=47675213https://news.ycombinator.com/item?id=47675213<![CDATA[Kubernetes Resource Optimization Strategies That Work in Production]]>Article URL: https://scaleops.com/blog/5-kubernetes-resource-optimization-strategies-that-work-in-production/

Comments URL: https://news.ycombinator.com/item?id=44900202

Points: 7

# Comments: 1

]]>Thu, 14 Aug 2025 13:33:30 +0000https://scaleops.com/blog/5-kubernetes-resource-optimization-strategies-that-work-in-production/timbilthttps://news.ycombinator.com/item?id=44900202https://news.ycombinator.com/item?id=44900202<![CDATA[New comment by timbilt in "Benchmarking GPT-5 on 400 real-world code reviews"]]>Yes, but in a case like this it's a neutral third-party running the benchmark. So there isn't a direct incentive for them to favor one lab over another.

With public benchmarks we're trusting the labs not to cheat. And it's easy to "cheat" accidentally - they actually need to make a serious effort to not contaminate the training data.

And there's massive incentives for the labs to cheat in order to get the hype going around their launch and justify their massive investments in training. It doesn't have to be the CEO who's directing it. Can even be one/a few researchers who are responsible for a specific area of model performance and are under tremendous pressure to deliver.

]]>Fri, 08 Aug 2025 06:35:16 +0000https://news.ycombinator.com/item?id=44834159timbilthttps://news.ycombinator.com/item?id=44834159https://news.ycombinator.com/item?id=44834159<![CDATA[New comment by timbilt in "Benchmarking GPT-5 on 400 real-world code reviews"]]>> Unlike many public benchmarks, the PR Benchmark is private, and its data is not publicly released. This ensures models haven’t seen it during training, making results fairer and more indicative of real-world generalization.

This is key.

Public benchmarks are essentially trust-based and the trust just isn't there.

]]>Fri, 08 Aug 2025 06:23:07 +0000https://news.ycombinator.com/item?id=44834088timbilthttps://news.ycombinator.com/item?id=44834088https://news.ycombinator.com/item?id=44834088<![CDATA[Google's Reverse Acquihire of Windsurf and the Future of AI Developer Tools]]>Article URL: https://www.qodo.ai/blog/googles-reverse-acquihire-of-windsurf-a-glimpse-into-the-future-of-ai-developer-tools/

Comments URL: https://news.ycombinator.com/item?id=44570021

Points: 14

# Comments: 5

]]>Tue, 15 Jul 2025 11:25:00 +0000https://www.qodo.ai/blog/googles-reverse-acquihire-of-windsurf-a-glimpse-into-the-future-of-ai-developer-tools/timbilthttps://news.ycombinator.com/item?id=44570021https://news.ycombinator.com/item?id=44570021<![CDATA[Making file encryption fast and secure for teams with advanced key management]]>Article URL: https://dropbox.tech/security/file-encryption-teams-advanced-key-management

Comments URL: https://news.ycombinator.com/item?id=44569986

Points: 2

# Comments: 0

]]>Tue, 15 Jul 2025 11:20:08 +0000https://dropbox.tech/security/file-encryption-teams-advanced-key-managementtimbilthttps://news.ycombinator.com/item?id=44569986https://news.ycombinator.com/item?id=44569986<![CDATA[Obesity drugs show promise for treating a new ailment: migraine]]>Article URL: https://www.nature.com/articles/d41586-025-01976-2

Comments URL: https://news.ycombinator.com/item?id=44377177

Points: 2

# Comments: 0

]]>Wed, 25 Jun 2025 13:35:16 +0000https://www.nature.com/articles/d41586-025-01976-2timbilthttps://news.ycombinator.com/item?id=44377177https://news.ycombinator.com/item?id=44377177<![CDATA[Slashing CI Costs at Uber]]>Article URL: https://www.uber.com/en-IL/blog/slashing-ci-costs-at-uber/

Comments URL: https://news.ycombinator.com/item?id=44375447

Points: 1

# Comments: 0

]]>Wed, 25 Jun 2025 09:57:45 +0000https://www.uber.com/en-IL/blog/slashing-ci-costs-at-uber/timbilthttps://news.ycombinator.com/item?id=44375447https://news.ycombinator.com/item?id=44375447<![CDATA[Detecting and Countering Malicious Uses of Claude]]>Article URL: https://www.anthropic.com/news/detecting-and-countering-malicious-uses-of-claude-march-2025

Comments URL: https://news.ycombinator.com/item?id=43774082

Points: 2

# Comments: 0

]]>Wed, 23 Apr 2025 16:52:28 +0000https://www.anthropic.com/news/detecting-and-countering-malicious-uses-of-claude-march-2025timbilthttps://news.ycombinator.com/item?id=43774082https://news.ycombinator.com/item?id=43774082<![CDATA[Uber's Journey to Ray on Kubernetes: Ray Setup]]>Article URL: https://www.uber.com/en-IL/blog/ubers-journey-to-ray-on-kubernetes-ray-setup/

Comments URL: https://news.ycombinator.com/item?id=43573769

Points: 1

# Comments: 0

]]>Thu, 03 Apr 2025 18:49:10 +0000https://www.uber.com/en-IL/blog/ubers-journey-to-ray-on-kubernetes-ray-setup/timbilthttps://news.ycombinator.com/item?id=43573769https://news.ycombinator.com/item?id=43573769<![CDATA[Why we chose LangGraph to build our coding agent]]>Article URL: https://www.qodo.ai/blog/why-we-chose-langgraph-to-build-our-coding-agent/

Comments URL: https://news.ycombinator.com/item?id=43459535

Points: 17

# Comments: 9

]]>Mon, 24 Mar 2025 11:04:27 +0000https://www.qodo.ai/blog/why-we-chose-langgraph-to-build-our-coding-agent/timbilthttps://news.ycombinator.com/item?id=43459535https://news.ycombinator.com/item?id=43459535<![CDATA[New comment by timbilt in "SOTA Code Retrieval with Efficient Code Embedding Models"]]>anyone else concerned that training models on synthetic, LLM-generated data might push us into a linguistic feedback loop? relying on LLM text for training could bias the next model towards even more overuse of words like "delve", "showcasing", and "underscores"...

]]>Mon, 03 Mar 2025 15:55:28 +0000https://news.ycombinator.com/item?id=43242945timbilthttps://news.ycombinator.com/item?id=43242945https://news.ycombinator.com/item?id=43242945<![CDATA[The Hidden Costs of Men's Social Isolation]]>Article URL: https://www.scientificamerican.com/article/the-hidden-costs-of-mens-social-isolation/

Comments URL: https://news.ycombinator.com/item?id=43049740

Points: 35

# Comments: 5

]]>Fri, 14 Feb 2025 16:00:10 +0000https://www.scientificamerican.com/article/the-hidden-costs-of-mens-social-isolation/timbilthttps://news.ycombinator.com/item?id=43049740https://news.ycombinator.com/item?id=43049740<![CDATA[Looking back at our Bug Bounty program in 2024]]>Article URL: https://engineering.fb.com/2025/02/13/security/looking-back-at-our-bug-bounty-program-in-2024/

Comments URL: https://news.ycombinator.com/item?id=43049715

Points: 1

# Comments: 0

]]>Fri, 14 Feb 2025 15:57:50 +0000https://engineering.fb.com/2025/02/13/security/looking-back-at-our-bug-bounty-program-in-2024/timbilthttps://news.ycombinator.com/item?id=43049715https://news.ycombinator.com/item?id=43049715<![CDATA[New comment by timbilt in "Scaling up test-time compute with latent reasoning: A recurrent depth approach"]]>Twitter thread about this by the author: https://x.com/jonasgeiping/status/1888985929727037514

]]>Mon, 10 Feb 2025 19:51:08 +0000https://news.ycombinator.com/item?id=43004425timbilthttps://news.ycombinator.com/item?id=43004425https://news.ycombinator.com/item?id=43004425<![CDATA[Scaling up test-time compute with latent reasoning: A recurrent depth approach]]>Article URL: https://arxiv.org/abs/2502.05171

Comments URL: https://news.ycombinator.com/item?id=43004416

Points: 149

# Comments: 44

]]>Mon, 10 Feb 2025 19:50:20 +0000https://arxiv.org/abs/2502.05171timbilthttps://news.ycombinator.com/item?id=43004416https://news.ycombinator.com/item?id=43004416<![CDATA[AI is accelerating scientific production, not progress]]>Article URL: https://twitter.com/random_walker/status/1888919433671843947

Comments URL: https://news.ycombinator.com/item?id=42999742

Points: 5

# Comments: 0

]]>Mon, 10 Feb 2025 13:05:00 +0000https://twitter.com/random_walker/status/1888919433671843947timbilthttps://news.ycombinator.com/item?id=42999742https://news.ycombinator.com/item?id=42999742<![CDATA[The Sport of Tuning Quantum Dot Arrays: QDarts Simulator]]>Article URL: https://www.quantum-machines.co/blog/the-sport-of-tuning-quantum-dot-arrays-how-the-qdarts-simulator-is-changing-the-game/

Comments URL: https://news.ycombinator.com/item?id=42923145

Points: 1

# Comments: 0

]]>Mon, 03 Feb 2025 21:07:30 +0000https://www.quantum-machines.co/blog/the-sport-of-tuning-quantum-dot-arrays-how-the-qdarts-simulator-is-changing-the-game/timbilthttps://news.ycombinator.com/item?id=42923145https://news.ycombinator.com/item?id=42923145<![CDATA[New comment by timbilt in "Solving key challenges in AI-assisted code reviews"]]>Until we get real-time learning to work in production, every AI tool feels like it's getting dumber over time. It goes very quick from "wow this is magic" to starting to notice all the little gaps. I think we have a fundamental expectation of intelligence to learn and when it doesn't, it just doesn't seem that smart

]]>Sun, 02 Feb 2025 09:07:52 +0000https://news.ycombinator.com/item?id=42907357timbilthttps://news.ycombinator.com/item?id=42907357https://news.ycombinator.com/item?id=42907357