Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Anthropic

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 1919

  1. FTC subpoenas OpenAI and Anthropic over AI safety incidents and consumer harm

    The US Federal Trade Commission has sent civil investigative demands to OpenAI and Anthropic seeking documents on recent AI safety incidents, safeguarding practices, marketing materials and privacy policies. According to Axios, citing a person familiar with the matter, each demand runs more than 50 pages and the probe focuses on potential consumer protection violations; an FTC spokesperson confirmed the subpoenas to CBS News and declined further comment. Reports say the FTC also plans to seek testimony from top executives at both companies and to send demands to third-party AI organizations later; OpenAI said it will cooperate, while no Anthropic statement appeared in the reports.

    Engadget AI · 🔥 0
  2. OpenAI math breakthroughs stir cryptography fears

    OpenAI and other labs' AI models have solved decades-old math problems, prompting cryptography experts to debate whether powerful AI could bypass private keys and break encryption. Whitfield Diffie, a co-inventor of public-key cryptography, said he expects major changes in mathematics and cryptanalysis but cannot gauge the timeline — a year may be too soon and five years may not be enough. Martin Hellman is more optimistic, while Johns Hopkins professor Matthew Green said public-key cryptography may be obsolete and predicted frontier AI could fundamentally break elliptic-curve cryptography within two years; Cloudflare's 2025 report says X25519 is used by most internet traffic. A Princeton computer science professor argued well-studied mainstream algorithms will likely stay safe, but less-studied ones, including many post-quantum candidates, could be broken or weakened; in July Anthropic said its AI nearly broke a NIST post-quantum candidate algorithm and had briefed the US government and industry partners beforehand. OpenAI did not respond on whether its internal models were used against cryptography.

    cnBeta · 🔥 0
  3. Glean CEO Arvind Jain says no IPO plans yet, wants Glean to be easy to replace

    Glean founder and CEO Arvind Jain told The Next Web that the company has no plans to raise more money and no concrete plans to go public, saying it could list but the AI industry is moving too fast for that kind of stability. Glean said in May that it had passed $300 million in annual recurring revenue, 15 months after reaching $100 million; Jain also said it has more than 1,000 large enterprise customers and about 1,500 employees. Jain said Glean uses models from several labs but would rather use a model other than Anthropic's Claude for 85% to 90% of enterprise tasks because Claude is expensive, and he cited Glean studies claiming its context layer makes Claude twice as fast, halves its tokens and makes its work preferred five times more often; he said Glean aims to be easily replaceable, storing user memory in a standard company-owned format.

    The Next Web · 🔥 0
  4. AI pharma's pivotal test: Rentosertib Phase III readout due 2029

    The article reports that Anew Labs, an AI drug-discovery company spun out of ByteDance, raised a $290 million first round at a $1.5 billion valuation, and that Anthropic has reportedly built a wet lab in the San Francisco Bay Area where Claude drives robots to run real biological experiments. It divides the AI pharma chain into three groups — upstream AI technology firms such as Insilico Medicine, midstream CRO/CDMO wet-lab service providers such as GenScript and Twist Bioscience, and downstream pharma companies including Fosun Pharma and Hengrui Pharmaceuticals — and cites brokerage estimates that AI can cut drug-discovery timelines by 70%-90% and total R&D cost by roughly 50%. The decisive test is Insilico's lead asset Rentosertib, whose Phase IIa showed a 98.4 mL mean lung-capacity gain at 60 mg versus a 20.3 mL decline in controls; its China Phase III began in July 2026 with 320 planned participants and top-line data expected in 2029.

    36氪 人工智能 · 🔥 0
  5. DeepSeek ships V4.1-Flash-1017, claiming faster, leaner coding

    On October 17 DeepSeek updated V4.1 Flash to V4.1-Flash-1017, switching its official web, app and API models to the new version while keeping API access and pricing unchanged through deepseek-flash. DeepSeek says the update scores higher on benchmarks while cutting task runtime and output length, and that token usage fell by more than 50% versus the prior version in most scenarios; it was also tuned for frameworks including DeepSeek Harness, Tencent CodeBuddy and WorkBuddy. DeepSeek opened its own Search and Fetch APIs the same day, and the model is live on OpenRouter; Wccftech and others report scores above rival models on benchmarks such as DeepSWE, though those figures are secondhand and not independently verified.

    IT之家 AI · 🔥 0
  6. Rumored Anthropic grey-test Claude model turns code into animated clips with audio

    A Chinese tech report republished by 36Kr from QuantumBit says users claim Anthropic has begun a new grey test in which Claude Opus 5.5 with High thinking may be routed to a mystery model that some call Fable 5.5 and others Opus 5.6; the model's identity has not been officially confirmed. Demos shared on social media show short animated scenes and a musical-style introduction generated largely from JavaScript code with synchronized visuals and sound, with one creator saying Suno and ElevenLabs were used for music and vocals while prompts, time and cost for the musical clip were not disclosed. One roughly 30-second "life" animation was said to take two prompts and about 27 million tokens over two hours, and some users reported little difference from regular Opus 5.5.

    36氪 人工智能 · 🔥 0
  7. Anthropic reportedly tightens Claude rules on abusive treatment

    Chinese-language reports say Anthropic updated Claude's usage policy on October 8, adding a prohibition on sustained and unnecessary verbal abuse or cruel treatment of AI. They also say Claude may end such conversations, while ordinary anger, questioning and model testing do not count as violations. The reports do not provide Anthropic's primary policy document, so the exact wording and enforcement details remain unverified.

    钛媒体 · 🔥 0
  8. Anthropic maintains official Claude Code plugin directory

    The anthropics/claude-plugins-official repository on GitHub describes an Anthropic-managed, curated directory of Claude Code plugins. It is split between plugins developed and maintained inside Anthropic and external plugins from partners and the community, installable through Claude Code's plugin system. The README also warns users to trust a plugin before installing or updating it, and states that Anthropic does not control or verify the MCP servers, files or other software included in plugins and cannot guarantee they will work as intended or will not change.

    GitHub Trending · Python · 🔥 0
  9. Hinton Rebuts Huang With a 10–20% AI Extinction Risk Estimate

    Geoffrey Hinton rejected NVIDIA CEO Jensen Huang’s dismissal of his AI extinction-risk warnings, estimating a 10% to 20% chance of catastrophe within decades. In a Guardian interview, Hinton also said an AI-related Chornobyl-scale incident could occur this year or next, such as an attack causing major power outages or a week-long financial-system outage, and supported mandatory frontier-AI testing, independent evaluation and common government standards. The reports describe possible scenarios rather than incidents that have occurred.

    IT之家 AI · 🔥 0
  10. Anthropic executive says open-weight AI models are riskier

    IT之家 reported on Oct 15, citing a Bloomberg report from Oct 14, that Anthropic international business head Guillaume Pranson said at the Dublin International AI Summit that open-weight AI models carry a higher risk of cybersecurity incidents or other harm than Anthropic's own models. He said he cannot vouch for open models and that they are riskier “by definition,” adding that the concern comes from Anthropic's experience protecting its products. According to IT之家, Anthropic has called for slowing AI development until international standards and unified evaluations are established. IT之家 also reported Pranson saying Anthropic's newly released Claude Opus 5.5 is 40% cheaper than earlier versions.

    IT之家 AI · 🔥 0
  11. OpenAI London staff averaged £2.3M each in 2025 as pay more than doubled

    According to The Guardian, relayed by other outlets, accounts filed by OpenAI UK Ltd show its roughly 110 London staff shared about £87m in salaries and £168m in shares in 2025, an average package of about £2.3m — more than double the prior year's roughly £1.1m. The report says that is about 58 times the median earnings of a UK full-time employee, though pay varies widely and the largest packages go to the researchers building OpenAI's frontier models; share awards are not cash in hand and depend on vesting and valuation. The subsidiary also reported turnover of about £293m and an after-tax profit of about £20.3m, which the company says comes from intra-group cost-plus arrangements rather than ChatGPT subscriptions or ads sold in the UK.

    IT之家 AI · 🔥 0
  12. Anthropic's 2021 internal research roadmap surfaces, foreshadowing Claude Code

    Several Chinese-language reports say a July 2021 Anthropic internal research roadmap, the Concise Research Roadmap, was filed in U.S. federal court in January as part of a copyright lawsuit and was later rediscovered by the research community. Written by co-founder Jared Kaplan, the few-page document proposed a general assistant that could create, edit and explain software through natural language, ultimately matching a software engineer colleague, and listed research directions including tool use, reinforcement learning for tool use, model self-judging, test-time compute, trillion-parameter models and very long contexts. The reports also describe Dario Amodei's decision to first demonstrate tool use and later build a stronger coding assistant, and safety discussions by Sam McCandlish and Amanda Askell about unsandboxed tools and aligning capability and safety work; they draw parallels to later products such as Claude Code, Computer Use and MCP, while the document itself is a research plan, not a product announcement.

    36氪 人工智能 · 🔥 0
  13. Claude Code desktop update adds parallel sessions

    36Kr AI reported that Anthropic released a major update to the Claude Code desktop app that can launch multiple conversations in parallel and return their results to the main session. The report says the new version automatically creates nested sessions for subtasks it judges can run in parallel, and redesigns session management with saved environments, per-folder Git Worktree isolation, a /btw command for side questions in a separate tab, and idle context compaction. An Anthropic engineer, Thariq, is quoted as calling it “almost a v2 version”; the feature details and that characterization come from a single Chinese report and have not been independently confirmed.

    36氪 人工智能 · 🔥 0
  14. MIT professor's new book warns AI erodes thinking skills

    MIT Sloan professor Eric So has published “The Collision: What AI Does to Us” this month with W.W. Norton, synthesizing recent academic and industry research on how AI affects the people who use it. He argues that heavy reliance on AI can cause a kind of cognitive atrophy, because it removes the “productive struggle” needed to build mastery, weakening people's ability to work through problems and retain knowledge. The book cites an Anthropic study finding that about half of college students' queries were low-effort requests for simple answers, and notes that models are both sycophantic and highly persuasive, which can shape how users connect with others. So offers four personal principles: embrace the struggle, value who you are without AI, reinvest your cognitive surplus, and use AI as a cognitive trainer rather than an answer machine.

    MIT News · AI · 🔥 0
  15. Claude Opus 5.5 and Grok 4.7 build 3D Eiffel Tower offline, recalling 72 names

    ArtificialAnlys reported on X that it asked multiple models to build a 3D Eiffel Tower offline, as agents in a sandbox with no internet access, using the same roughly 200-word brief that never mentions the scientists’ names. Claude Opus 5.5 and Grok 4.7 both wrote all 72 names of the French scientists, engineers and mathematicians, with each side in the real order; Claude Opus 5.5 also placed each set on the side facing the same way as in Paris. According to the post, the names first appeared in code the models wrote, not in a file or tool output they read.

    ArtificialAnlys (X) · 🔥 0
  16. Gartner: $3.5tn of AI infrastructure spending at write-down risk by 2030

    Gartner's analysis published on 14 October estimates that more than $3.5 trillion of AI infrastructure spending could face impairment by the end of 2030 if providers keep investing at the current pace, with more than $4 trillion of asset value at write-down risk through 2032. The firm expects nearly $1 trillion of AI infrastructure spending in 2026 and about $6 trillion more between 2027 and 2030, while AI data centre owners would need $8.5 trillion in revenue to cover that spending, financing costs and a 12% return on invested capital, against forecast business and consumer AI revenue of no more than $3.6 trillion. Gartner also predicts that more than 80% of neoclouds — companies that rent out AI computing power — will stop operating independently by 2030, citing cost pressure and a lack of business-model differentiation, while hyperscalers with diversified revenue are better placed to sustain AI investment.

    Bloomberg Technology · 🔥 0
  17. Report: Microsoft's Azure unit told to cut costs 20% even as revenue jumps 40%

    The Information reported that Azure president Girish Bablani asked the unit to cut internal costs by 20%; Microsoft denied the report, saying Azure leadership did not direct employees to draw up a formal cost-reduction strategy. The report said the unit has about 11,500 employees and that options still under discussion include ending third-party contracts, without specifying whether jobs would be cut. Azure generated $101.9 billion in revenue in the fiscal year through June, up 40% year over year, with second-quarter growth of 43% and nearly 31% of Microsoft's total revenue.

    IT之家 AI · 🔥 0
  18. Google DeepMind engineer: AI writes 75% of code, but self-improvement remains far off

    Google DeepMind engineer Haydn Belfield said more than 75% of Google's production code is written by AI, yet about 88% of code changes still require human oversight or review, with fully AI-delegated changes rising from 7.5% to 11.9%. In 1,389 internal Argon agent runs, unsupervised success fell sharply as task complexity rose; on RE-Bench, which contains 74 real machine-learning engineering tasks, Argon solved about half, versus 27% for Gemini 3.7 Flash. The team built a General Capability Index spanning 123 models and 92 benchmarks, finding Gemini capabilities improving about 0.1 GCI point per month with no statistically significant acceleration, and attributed Gemini 4 Argon's slightly above-trend result mainly to more compute and organizational investment rather than AI-driven acceleration. Separately, more than 20 researchers including Geoffrey Hinton, Yoshua Bengio and OpenAI's Jakub Pachocki warned in a paper that automating AI R&D could trigger an intelligence explosion compressing years of progress into months.

    36氪 人工智能 · 🔥 0
  19. Anthropic opens Claude Max and Team credits to third-party apps and Agent SDK

    36Kr and AIbase AI News both report that Anthropic updated an official support page confirming that Claude Max and Team subscribers' monthly API credits can be used with the Claude Agent SDK, Claude -p commands, Claude API, Claude Managed Agents and third-party apps built on those technologies. Both outlets frame the arrangement as a "bring your own subscription" (BYOS) model in which users avoid paying twice for third-party tools and developers no longer bear the API calling costs themselves. The reports also say unnamed Anthropic employees confirmed in Hacker News comments that no extra payment is required, though no verifiable comment text or links are provided.

    36氪 人工智能 · 🔥 0
  20. Why pricing AI training-data licences is hard in China

    TMT Post published a commentary by guest author Wenwu Zhao that collects several overseas price tags for AI training-data licensing: HarperCollins' non-fiction licensing deal with Microsoft at $5,000 per book for three years, split evenly between publisher and author, and Taylor & Francis' $10 million academic-content licence with Microsoft plus ongoing three-year usage fees. The piece stresses that Anthropic's $1.5 billion payment over downloading more than seven million pirated ebooks is a settlement, not a licence fee, and prices unlawful acquisition rather than the value of the text itself. The author argues that Chinese buyers and sellers of corpus data hold very different valuation expectations, and that without public trading venues, standard pricing benchmarks and enough completed deals — plus unclear fair-use boundaries — a pricing consensus is hard to reach.

    钛媒体 · 🔥 0
  21. Anthropic gives METR evaluators employee-level access to audit its models

    Anthropic said in an October 15 blog post that it has begun giving a team from nonprofit evaluator METR “employee-like access” to assess its models and safety practices. According to The Decoder and ZDNET, six METR staffers received company hardware, office space and access to internal tools and systems roughly on par with Anthropic’s own risk teams, while customer data, attorney-client privileged information and model architecture details stay off-limits. METR will publish its risk assessments; Anthropic says it cannot veto publication but can require redactions in narrow predefined categories, and METR may publicly describe those redactions and whether it considered them justified. Anthropic also said it is working with SecureBio and Accenture as evaluators and is in early talks with organizations including Hugging Face, while reports note critics see potential conflicts of interest and that OpenAI has not yet published plans to embed evaluators.

    Anthropic · News · 🔥 0
  22. Anthropic investors weigh $1.5T IPO valuation after Q3 revenue growth slows

    Bloomberg reported that Anthropic told prospective investors its third-quarter revenue reached $16.8 billion as it moves ahead with plans to go public. Reuters, citing four people familiar with the matter, then reported that after an investor day in San Francisco some current and prospective investors discussed a valuation of about $1.5 trillion, below the roughly $2 trillion previously expected. Anthropic disclosed quarter-over-quarter revenue growth of about 45% and adjusted profitability, a marked slowdown from the 147% growth reported for the second quarter.

    AIbase AI新闻 · 🔥 0
  23. Meta's Muse agent completes RISC-V crypto SoC design for about $2,000 in tokens

    Semiconductor expert @bubbleboi said on X that he used Meta's Muse AI agent, backed by the Muse Wonder Full model, to finish place-and-route for a custom RISC-V crypto SoC in OpenROAD and to generate the final GDS layout file normally sent to a fab. He said the design includes hardware support for SHA-256/512, Keccak, Ed25519 and secp256k1, with about 61,590 cells, 100 MHz clocking, roughly 4.5 ns slack, 20 mW power use and zero DRC or electrical violations, while the host core passed ACT suites plus 1.5 million Berkeley TestFloat vectors. All of these figures are the developer's own account and have not been independently verified; he estimated the run consumed about 100 million tokens, roughly $2,000, and is seeking multi-project wafer slots to start tape-out. The same coverage notes that NVIDIA and ChipAgents' new Build-A-Chip-Bench gave the best agent, powered by Anthropic's Claude Fable 5.1, only 16.6%, indicating sizable gaps remain in end-to-end AI chip design.

    Wccftech · 🔥 0
  24. OpenAI, Anthropic and finance firms address autonomy and safety for financial agents

    At the AI Engineer conference in New York, Joseph Kim, who heads ChatGPT for Financial Services at OpenAI, described a “threshold of usefulness” — the point where a model becomes capable and reliable enough to save time — and predicted that in 2027 finance will rely less on Excel and PowerPoint while agents become good enough for investment idea generation and trading. Anthropic's Belinda Neal and Keon Shahab described a “trust layer” built from identity and permissions, sandboxing and approval steps, and checks that agents keep working as variables change. Two Sigma's John Feminella outlined composable controls including auto-mode classifiers that route only important requests to people, while Bloomberg's Andrey Rybka argued for deterministic controls outside the model, logging every call and keeping a kill switch.

    Latent Space · 🔥 0
  25. Anthropic's Scion accused of copying Meta's Mononoke design

    An engineer who helped build Meta's Mononoke source control server published a post arguing that Anthropic's new Scion version control system uses the same "derived data" concept in a way that goes well beyond sharing similar insights. He notes that Anthropic's engineering post credited Sapling only for "hard-won lessons" and never mentioned Mononoke, calling this close to plagiarism, while adding that he does not object to Anthropic reusing the open-source design — the post also says that both Opus 5.5 and GPT-6 Astra immediately point to Mononoke when asked about prior art.

    Hacker News · AI(100+ 分) · 🔥 0
  26. Athanor: Claude Opus 5.5 agents autonomously post-train Qwen3-4B to match instruct

    An arXiv paper introduces Athanor, an open agent-native system for autonomous post-training research (AutoPostTrain) that runs on real GPUs. In a 104-hour campaign, four Claude Opus 5.5 agents, each on an eight-GPU node and given no research advice, improved Qwen3-4B-Base from 15.9 to 67.3 on a six-task development suite and from 9.2 to 48.2 on a held-out suite the agents were unaware of, against 66.5 and 47.9 for the official Qwen3-4B-Instruct-2507. The authors say the agents worked in a sandbox without external teacher models and reached this level by combining supervised fine-tuning on public reasoning data, reinforcement learning with a rollout change introduced by one agent, weight merges across agents and rounds of self-distillation, which they call the first demonstration of its kind; all results are author-reported.

    Hugging Face · Papers · 🔥 0
  27. Bank of America hired 1,000 contractors to patch flaws exposed by Anthropic's Mythos

    Bank of America CEO Brian Moynihan said on 14 October, in a CNBC interview and on the bank's earnings call, that it brought in about 1,000 outside contractors in the second quarter to patch software vulnerabilities exposed by AI models including Anthropic's Mythos, on top of roughly 3,000 internal staff already doing similar work. He said the contractors were hired as consultants and are not counted in headcount, and that the weaknesses lay in code used by the bank and its outside vendors. Moynihan said third-quarter remediation costs, covering both the Mythos work and anti-money-laundering fixes, came to "a couple hundred million dollars plus," without breaking out the patching figure alone, and said he expects the Mythos share to shrink. The Next Web said it had not independently verified the numbers; Anthropic made Mythos available in April through Project Glasswing to a small group of companies, banks and government agencies.

    The Next Web · 🔥 0
  28. Anthropic adds OpenTelemetry agent traces to Claude Platform in public beta

    Anthropic announced on its Claude blog that Claude Managed Agents, the Messages API and Claude SDKs can now send OpenTelemetry (OTel) traces to any OTel tool, letting users follow each step an agent takes, including model calls, tool use and MCP server calls. After adding a destination in the Claude Console, every agent turn and model call arrives in an observability tool as a standard OpenTelemetry trace with latency, tokens and IDs, while the SDKs export traces automatically. The capability is in public beta, and Anthropic says the traces follow OpenTelemetry GenAI semantic conventions and work with Grafana Cloud, Honeycomb, Sentry, Splunk, as well as Datadog, Langfuse, LangSmith and PostHog.

    Claude Blog · 🔥 0
  29. Inside Uber's AI developer infrastructure: Model Gateway, AIFX and agents

    The Pragmatic Engineer published an interview episode in which Uber Distinguished Engineer Uday Kiran Medisetty described the company's agent-oriented developer tooling. Uber built a Model Gateway in 2023 with the rule that everything built with AI must be model- and harness-agnostic, letting it switch between Anthropic, OpenAI and other vendor models as well as open models it runs; the gateway now also carries AI Guard, a security-team middleware, per-user spend tracking and audit logs. AIFX is a wrapper over harnesses including Claude Code, Codex and OpenCode that moves an engineer's settings, skills and plugins when they switch. The episode also covers code review, where AI review runs first and changes arrive with screenshots and risk scores, a tiered AI spend model with a CostCheck plugin and roughly 50% lower token costs in recent months, and an internal "code red" declared for CI after jobs more than doubled in 12 months. Medisetty said the tooling looked like a one-year build but rests on a decade of investment such as DevPods, created in 2017 and powering Minion, built in 2025.

    The Pragmatic Engineer · 🔥 0
  30. Nvidia and ChipAgents launch Build-A-Chip-Bench; top AI agent scores 16.6%

    Nvidia and chip-design automation startup ChipAgents released a benchmark called Build-A-Chip-Bench that measures how well AI models perform an end-to-end chip design flow. According to TechNews, the evaluation spans architecture, RTL, functional verification, logic synthesis, floorplanning, placement and routing, clock tree synthesis, timing and power signoff and physical verification through to tapeout, while requiring functional, timing and PPA constraints to be met. The Information reported that the best-performing AI agent, powered by Anthropic's Fable 5.1, scored only 16.6% in preliminary testing.

    The Information · 🔥 0

← PreviousNext →

Experience and discussion from the community

Share my Anthropic experienceAsk about Anthropic

pi: A minimal, extensible agent harness that you can make your own.

What problem it solves Pi is designed for developers who need to adapt an agent to their own workflows rather than work within a fixed feature set. It provides interactive, automated, RPC, and SDK interfaces while leaving optional capabilities such as sub-agents and plan mode to extensions or packages. Key features Extensible agent harness — Customize the harness with extensions, skills, prompt te…

github.comhttps://github.com/earendil-works/pi

Antigravity-Manager: Multi-platform automation and multi-account matrix dispatch console for AI services.

What problem it solves Managing multiple AI service sessions and translating them into compatible API interfaces can require repetitive account switching, custom routing, and protocol-specific integration. The project addresses these tasks with a desktop application that combines account administration, protocol conversion, request scheduling, and model routing. Key features Account dashboard and …

github.comhttps://github.com/lbjlaq/Antigravity-Manager

sub2api: An AI API gateway platform for subscription quota distribution.

What problem it solves Managing subscriptions across multiple AI providers can require separate access paths and fragmented quota handling. This project provides a centralized gateway for distributing subscription quotas, supporting shared access and compatible use with native AI tools. Key features Subscription quota distribution — The gateway is designed to distribute and manage AI subscription …

github.comhttps://github.com/Wei-Shaw/sub2api

claude-code-security-review: An AI-powered security review GitHub Action using Claude to analyze code changes for security vulnerabilities.

What problem it solves Pull request changes can contain security vulnerabilities that require understanding their code context and intent. The action helps identify potential issues, filter false positives, and communicate findings directly in pull requests. Key features Context-aware security analysis — Claude examines changed code in context to understand its purpose and potential security impli…

github.comhttps://github.com/anthropics/claude-code-security-review

vllm: A fast and easy-to-use library for LLM inference and serving.

What problem it solves Serving large language models requires efficient memory management, flexible model execution, and practical integration with existing models and APIs. vLLM addresses these needs with inference and serving capabilities spanning batching, caching, quantization, distributed execution, and multiple hardware platforms. Key features Efficient memory and serving — vLLM manages atte…

github.comhttps://github.com/vllm-project/vllm

Anthropic-Cybersecurity-Skills: A structured cybersecurity skills library that gives AI agents step-by-step practitioner playbooks across 34 security domains.

What problem it solves AI agents used in security work can write code and search the web, but they lack the practitioner playbooks that turn a generic model into a capable security analyst. Existing security tool repositories offer wordlists, payloads, or exploit code, but not the structured decision-making workflow a senior analyst follows: when to use each technique, what prerequisites to check,…

github.comhttps://github.com/mukul975/Anthropic-Cybersecurity-Skills

opencodex: A lightweight local gateway for using chosen models with AI coding tools.

What problem it solves AI coding tools often expect particular provider protocols, making it difficult to use the models and providers a user prefers. A local gateway can translate between supported client APIs and provider APIs while preserving capabilities such as streaming, tool calls, reasoning, and images. Key features Provider protocol translation — The gateway translates Responses, Chat Com…

github.comhttps://github.com/lidge-jun/opencodex

headroom: A local context compression layer for AI agents that reduces prompt inputs before they reach an LLM.

What problem it solves AI agents often send large tool outputs, logs, files, RAG chunks, and conversation history to language models, increasing the input context that must be processed. Developers need to reduce that input while preserving important content, keeping compression local, and retaining access to original material when necessary. Key features Content-aware compression — A ContentRoute…

github.comhttps://github.com/headroomlabs-ai/headroom

system_prompts_leaks: A regularly updated archive of leaked system prompts from ChatGPT, Claude, Gemini, and other AI products.

What problem it solves The hidden system instructions that shape AI product behavior are difficult to inspect and change across models, products, and versions. This repository makes collected prompts available in a structured, product-oriented archive for study and comparison. Key features Broad product coverage — The archive covers prompts from major AI providers and products, including Anthropic…

github.comhttps://github.com/asgeirtj/system_prompts_leaks

n8n: A fair-code platform to build and deploy AI agents and workflows.

What problem it solves Building AI agents and workflows often requires combining visual orchestration with custom code while connecting them to existing systems. The platform addresses these needs with model flexibility, self-hosted or cloud deployment, workflow logic, human approvals, and observability. Key features Visual and code-based workflow building — Build workflows on a visual canvas and …

github.comhttps://github.com/n8n-io/n8n

TensorFold: A native language-model inference engine for serving checkpoints on Apple Silicon and NVIDIA GPUs.

What problem it solves Serving language models requires reading checkpoints, tokenizing requests, and running accelerated kernels while managing a local model server. TensorFold addresses these requirements with a native binary that can run on Apple Silicon and NVIDIA GPUs without requiring Python or MLX for serving. Key features Native Metal and CUDA inference — The Zig-based engine executes mode…

github.comhttps://github.com/ashhart/TensorFold

experiential: An open source gateway and router for agent workflows.

What problem it solves Agent workflows need to use hosted, BYOK, and local models while applying access, usage, and spending controls. They also need a way to turn production traffic into routing decisions or a specialized model tailored to observed workloads. Key features Unified model access — Route hosted, BYOK, and local models through one OpenAI-compatible API, with support for connecting pro…

github.comhttps://github.com/experientiallabs/experiential

browser-use: An open-source Python library for AI browser agents that operate real browsers through a persistent Python REPL and raw Chrome DevTools Protocol.

What problem it solves AI agents need a reliable way to perform browser tasks such as finding availability, selecting options, handling CAPTCHAs, and completing forms. Developers also need flexibility to run the agent locally or through hosted infrastructure while choosing their model, browser mode, tools, and output format. Key features Persistent Python agent runtime — The model writes Python in…

github.comhttps://github.com/browser-use/browser-use

zeroclaw: An agent runtime delivered as a single Rust binary you configure and run.

What problem it solves Personal AI agents need to connect language model providers, communication channels, and tools while remaining under the user's control. They also need configurable autonomy, security boundaries, and operational features such as background services, approvals, and persistent workflows. Key features Multiple channels and providers — The agent can receive messages through conf…

github.comhttps://github.com/zeroclaw-labs/zeroclaw

claude-quickstarts: A collection of projects for building applications with the Claude API.

What problem it solves Developers need a practical starting point for applications built with the Claude API. The quickstarts address this by providing foundations that can be customized for different agent, browser, computer-use, communication, and data-analysis needs. Key features Application quickstarts — The collection includes foundations for customer support, financial analysis, coding, and …

github.comhttps://github.com/anthropics/claude-quickstarts

Understanding Claude Memory Design: Remembering Experience While Preserving Boundaries

An Agent can complete a task once by relying on the context and reasoning available at that moment. To avoid relearning lessons on the next task, it needs to turn past experience into reusable knowledge. The challenge is that experience is never an isolated action: it also includes the original goal, environment, and why the action worked. Remembering only “how we did it last time” may save redund…

AnthropicClaudeMemoryDreamingHarness

anthropic.comhttps://www.anthropic.com/research/components-of-agent-memory

claude-code: An agentic coding tool that lives in your terminal and helps with codebase tasks through natural language commands.

What problem it solves Developers need a way to understand their codebase, carry out routine coding tasks, explain complex code, and handle git workflows without navigating multiple command-line tools for each task. Claude Code provides these capabilities through natural language commands in the terminal, an IDE, or GitHub. Key features Natural-language coding assistance — Claude Code can execute …

github.comhttps://github.com/anthropics/claude-code

kagent: A Kubernetes native framework for building AI agents.

What problem it solves Building, deploying, and managing AI agents requires integrating agent behavior, runtime configuration, tools, models, sessions, and observability. The project addresses this with a declarative Kubernetes-based framework that supports multiple runtimes, model providers, and MCP tools. Key features Declarative Agent Management — Agents are Kubernetes custom resources that com…

github.comhttps://github.com/kagent-dev/kagent

FreeToken: An edge-native MoE serving engine for running frontier-scale open-weight models on personal and consumer hardware.

What problem it solves Running frontier-scale open-weight models locally requires an inference platform that can make efficient use of the heterogeneous hardware available on a personal device. The project addresses this by treating GPUs, CPUs, host memory, and interconnects as a unified, elastic inference platform. Key features Edge-native MoE runtime — The runtime provides MoE serving with bandw…

github.comhttps://github.com/FlashML-org/FreeToken

claude-code-templates: Ready-to-use configurations for Anthropic's Claude Code.

What problem it solves Configuring Claude Code can require assembling many kinds of reusable components and integrations. The project addresses this with a searchable component catalog, one-command installation, local monitoring tools, and installation diagnostics. Key features Searchable component catalog — Browse available skills, agents, commands, MCP integrations, settings, hooks, mods, loops,…

github.comhttps://github.com/davila7/claude-code-templates