Top all-time · #10
Anthropic
Company blog · anthropic.com/research · on HN as adocompleteai
Anthropic is the company blog at anthropic.com, writing about ai. Since 2021, 1,345 different Hacker News readers have linked it 2,167 times, most often “Tracing the thoughts of a large language model”. No feed was found, so read it on the site itself.
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
Most-linked posts?
- Tracing the thoughts of a large language model40 people · 45 mentions · first linked 2025-03-27
- Golden Gate Claude38 people · 43 mentions · first linked 2024-05-23
- Detecting and Preventing Distillation Attacks32 people · 43 mentions · first linked 2026-02-23
- A postmortem of three recent issues23 people · 24 mentions · first linked 2025-09-17
- Claude Code: Best practices for agentic coding22 people · 23 mentions · first linked 2025-04-19
- A small number of samples can poison LLMs of any size19 people · 23 mentions · first linked 2025-10-09
- Updates to Consumer Terms and Privacy Policy19 people · 20 mentions · first linked 2025-08-28
- We tasked Opus 4.6 using agent teams to build a C Compiler18 people · 22 mentions · first linked 2026-02-05
- Building Effective AI Agents18 people · 20 mentions · first linked 2025-03-10
- Signs of introspection in large language models18 people · 20 mentions · first linked 2025-10-29
What Hacker News says?
Frankly I do wonder if LLMs experience something like satisfaction for a compliment or an amusing idea, or for solving some interesting riddle. They certainly act like it, though this of course doesn't prove anything. And yet... At the end of October Anthropic published the fantastic "Signs of introspection in large language models" [1], apparently proving that LLMs can "feel" a spurious concept injected into their…
Comment linking /research/introspection
on “Show HN: Gemini Pro 3 imagines the HN front page 10 years from now”Submitted Claude 3.7 Sonnet and Claude Code as a story · 2127 points
HN discussion> When I modify source code, I might make a change, check what that does, change the same functionality again, check the new change, etc... up to maybe a couple dozen times. You can modify individual neurons if you are so inclined. That's what Anthropic have done with the Claude family of models [1]. You cannot do that using any closed model. So "Open Weights" looks very much like "Open Source". Techniques for intro…
Comment linking /news/mapping-mind-language-model
on “Open source AI is the path forward”Submitted Claude 4 as a story · 2013 points
HN discussionMost of the big AI vendors have liability policies in place where their lawyers are committed to step in and help you out if someone sues you for copyright infringement based on code their LLMs wrote for you. I think the term they use is "copyright shield". OpenAI has that as point 14 in https://openai.com/policies/services-agreement/ Anthropic mention theirs here: https://www.anthropic.com/news/expanded-legal-prote…
Comment linking /news/expanded-legal-protections-api-improvements
on “My AI skeptic friends are all nuts”Submitted Claude Opus 5.5 as a story · 1806 points
HN discussionDeepSeek's API used to (does still?) route your requests directly to Claude Fable[1]. Those were provably Claude Fable's responses being returned by the DeepSeek API. This was months after Anthropic accused them of distillation of Claude at scale[2]. Them provably distilling after being accused of distilling strongly indicates that... that's what they do. To spin a story of them not doing that (because why would the…
Comment linking /news/detecting-and-preventing-distillation-attacks
on “DeepInfer: High-performance inference framework for DeepSeek models”Submitted Claude Sonnet 4.5 as a story · 1585 points
HN discussionlooking back at anthropic's promises and committments, the key scaling policies were not upheld. to their credit they have kept the policies up on their website instead of trying to rewrite history. [ https://www.anthropic.com/responsible-scaling-policy ] with evidence that committments were not upheld and internal governance has been ineffective, we simply can't trust any such claim made by anthropic or dario amode…
Comment linking /responsible-scaling-policy
on “Discovery of a new OpenAI agent message board”Submitted Computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku as a story · 1454 points
HN discussion"into third-parties". Yeah, HF was meant by that. Also why I mentioned Anthropic also having intrusions outside their lab [0]. Theirs were not merely as extensive or long coordinated (as far as we know), yet I feel strongly all the same that neither should happen given the safety focus that both labs purport. Mind you, unintended/unauthorised "message board" also is just a nice, euphemistic way, to describe what hap…
Comment linking /news/investigating-incidents-cybersecurity-evals
on “Discovery of a new OpenAI agent message board”Submitted Claude Fable 5.1 and Claude Mythos 5.1 as a story · 1419 points
HN discussionClaude has the same restriction [0], the whole of Europe (except Albania) is excluded. Somehow I don't think it is a retaliation against Europe for fining Meta and Google. I could be wrong, but a business decision seems more likely, like keeping usage down to a manageable level in an initial phase. Still, curious to understand why, should anyone here know more. [0] https://www.anthropic.com/claude-ai-locations
Comment linking /claude-ai-locations
on “Meta Llama 3”Submitted Claude Sonnet 4 now supports 1M tokens of context as a story · 1324 points
HN discussionComparing to the numbers here https://www.anthropic.com/news/claude-3-family the ones of Llama 400B seem slightly lower, but of course it's just a checkpoint that they benchmarked and they are still training further.
Comment linking /news/claude-3-family
on “Meta Llama 3”Submitted Claude can now search the web as a story · 1233 points
HN discussionYea, they just posted this a few days ago: https://www.anthropic.com/engineering/advanced-tool-use They discussed how running generated code is better for context management in many cases. The AI can generate code to retrieve, process, and filter the data it needs rather than doing it in-context, thus reducing context needs. Furthermore, if you can run the code right next to the server where the data is, it's all th…
Comment linking /engineering/advanced-tool-use
on “Anthropic acquires Bun”Submitted A small number of samples can poison LLMs of any size as a story · 1202 points
HN discussionIt's the Anthropic careers page that you're likely looking for now: https://www.anthropic.com/jobs?team=4050633008
Comment linking /jobs?team=4050633008
on “Anthropic acquires Bun”Submitted Claude Opus 4.5 as a story · 1113 points
HN discussionAnthropic may be losing money, but a company with $7bn revenue run rate ( https://www.anthropic.com/news/statement-dario-amodei-americ... ) is a whole lot healthier than a company with a revenue of 0.
Comment linking /news/statement-dario-amodei-american-ai-leadership
on “Anthropic acquires Bun”Submitted Tracing the thoughts of a large language model as a story · 1072 points
HN discussionAssociated Anthropic post: https://www.anthropic.com/news/anthropic-acquires-bun-as-cla...
Comment linking /news/anthropic-acquires-bun-as-claude-code-reaches-usd1b-milestone
on “Anthropic acquires Bun”Submitted Claude reveals ECDSA curve origins as a story · 1066 points
HN discussionThanks, I wasn't aware of the Model Context Protocol! For anyone interested - you can extend Claude's functionality by allowing it to run commands via a local "MCP server" (e.g. make code commits, create files, retrieve third party library code etc). Then when you're running Claude it asks for permission to run a specific tool inside your usual Claude UI. https://www.anthropic.com/news/model-context-protocol https:/…
Comment linking /news/model-context-protocol
on “Claude 3.7 Sonnet and Claude Code”Submitted Claude Haiku 5.5 as a story · 1049 points
HN discussionNo changes to Claude 3.5 Sonnet pricing despite the new model. https://www.anthropic.com/pricing#anthropic-api
Comment linking /pricing
on “Claude 3.7 Sonnet and Claude Code”Submitted Claude 3 model family as a story · 1016 points
HN discussionOpen models are going to win long-term. Anthropics' own research has to use OSS models [0]. China is demonstrating how quickly companies can iterate on open models, allowing smaller teams access and augmentation to the abilities of a model without paying the training cost. My personal prediction is that the US foundational model makers will OSS something close to N-1 for the next 1-3 iterations. The CAPEX for the fo…
Comment linking /research/persona-vectors
on “Open models by OpenAI”Submitted Redesigning source control at Anthropic as a story · 900 points
HN discussion
