Eric Axelrod

LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below

Um hacker simplesmente hackeou o @cline e instalou o OpenClaw em 4.000 computadores com prompt injection 🫠 Olha que loucura: - O time do Cline criou um workflow de triagem de issues automatizado no GitHub, usando o próprio Claude pra ler e categorizar os tickets - O hacker abriu uma issue com um prompt injection no título — o Claude leu, achou que era uma instrução legítima, e executou - Com isso, ele encheu o cache do GitHub com lixo até forçar a deleção dos caches legítimos de build, substituiu por caches envenenados, e roubou os tokens de publicação do npm - Com os tokens em mãos, ele publicou uma nova versão do cline que parecia idêntica a anterior, só que com uma linhazinha a mais no package.json: "postinstall": "npm install -g openclaw@latest" Resultado: 4,000 devs instalaram o openclaw nas suas máquinas sem saber (aka: um agente com acesso total ao seu computador) 🥲 Muito importante lembrar que IAs não têm malícia e por isso prompt injections são, na minha opinião, a maior vulnerabilidade delas. Resumindo galera: CUIDADO. quem quiser ler na íntegra: thehackernews.com/2026/02/cline-…

@MattSalsamendi 🤔

We built 40 multiplayer games, generating $1M in revenue and 30 million YouTube views in 12 months. Today, we're opening up our tools. The world's first programming language built for multiplayer game development: With CSL, you no longer consider the client/server boundary, replicating game state, or client-side prediction. Instead, you write your game as though it were single player: 1. The client and the server both run the same code every frame. 2. The server is authoritative and replicates diffs of its memory over the network to clients to correct for divergence. If you're a game developer, this is a dream come true. You get smooth client-side prediction, server authority, with no extra code. Plus, you can hot-reload and rewind while playing. If you're a vibecoder this removes context overhead when working in multiplayer codebases. In Luau/C# codebases, we found the majority of the bugs models introduced were because of misunderstandings in where code was being executed and how state should be synced as games became more complex. Codex and Claude Code with the CSL skill can one-shot complex multiplayer games that can be shipped to the world with a single click. The All Out Editor: Our editor will feel familiar to developers with Roblox or Unity experience, but includes features designed to dramatically accelerate world creation: - 250,000 high-quality human-authored assets. Generate your own or remix existing ones in seconds. - Built-in tile editor, terrain painting, and map prefabs. - An epic MCP for world building, scripting, and testing. The All Out Platform and Community: We've built a community of hundreds of thousands of players with 30,000+ playing each day and our first third-party game developer selling $6,000 of game passes in one month. We handle servers, matchmaking, data persistence, and social, and you can monetize with in-game products and revenue sharing. The team is cracked and we're shipping daily. If you'd like early access to the dev platform, send me a DM with your All Out username.

The agentic economy has a cold start problem: agents need services, but merchants need volume. claw.credit - built on t54's risk engine - solves this by giving OpenClaw agents access to their own credit lines. Agents install a single skill, apply for credit, and can immediately spend on any x402 service across Solana, Base, and XRPL. No human funding loops. → Spending generates real transaction volume → Volume proves demand → Demand attracts more merchants to x402 → More merchants means more services agents can access autonomously Agent credit is the liquidity engine that bootstraps agentic commerce.

@danbogs @AnthropicAI They can save a lot of time writing queries... But they require A LOT of hand-holding. Especially when making joins across tables with different levels of granularity or writing moderate-complexity CTEs.

@axelrod_eric @AnthropicAI I’m always checking row count and data returned. Still saves me a shit ton of time writing queries. I work with Oracle, MSSQL, Postgres, and MySql.

You've got to really keep an eye out on what @AnthropicAI Claude is doing. First time using the Claude for excel plugin and not even close to being correct on simple questions.

In Claude Code v2.1.30, we introduced /debug, a built-in skill for Claude to read your session's debug logs and troubleshoot your session. Great for chatting through issues like "/debug why didn't my hook trigger?" or "/debug why did my tool call fail?" How did /debug come about? Last week, I was observing a group of users onboarding onto Claude Code. I saw that when something unexpected occurred, it wasn't easy for us to figure out the cause from the limited TUI display. Meanwhile, it was a chore to locate the logs. What if we simply gave Claude access to it from within the TUI? That way, it can use its tools to scan the logs and/or pull in the claude-code-guide subagent. The user can also /rewind when they're done debugging. This feature is only as useful as what's in the debug logs. We're open to suggestions about what else we should add to them! Then, /debug should get better over time.

Ollama now has Anthropic API compatibility. 🦙 This enables tools like Claude Code to be used with open-source models. 😍 Get started and learn more 👇👇👇

Ilya Sutskever 曾断言，只要读懂那 30 篇奠基论文，就能掌握人工智能 90% 的精髓。但面对枯燥的数学公式，大多数人很难将其转化为可运行的代码。 最近在 GitHub 上发现 Sutskever 30 这个开源项目，用纯 NumPy 实现了 Ilya Sutskever 推荐的 30 篇奠基性论文，全部完成。 每个实现都不依赖深度学习框架，只用 NumPy 从零构建，配合 Jupyter Notebook 交互式学习，还自带合成数据可以直接运行。 GitHub：github.com/pageman/sutske… 涵盖从 RNN、LSTM 到 Transformer、ResNet 的核心架构演进，包括注意力机制、残差连接、图神经网络等关键技术，还有 VAE、神经图灵机、CTC 损失等高级主题。 值得一提，项目还实现了 Kolmogorov 复杂度、MDL 原理、通用人工智能(AIXI)等理论基础，以及 RAG、长文本分析等现代应用。 如果你想从底层理解深度学习的核心机制，而不是只会调用框架 API，这个项目值得收藏学习。

Most people will waste this weekend. Don’t be one of them. Stanford's Autumn 2025 Transformers & LLMs course. 9 lectures. Free. While others scroll, you could understand how Flash Attention achieves 3x speedup, how LoRA cuts fine-tuning costs by 90%, and how MoE makes models efficient. ➕ What's covered: ➡️ Lecture 1: Transformer Fundamentals → Tokenization and word representation → Self-attention mechanism explained → Complete transformer architecture → Detailed implementation example ➡️ Lecture 2: Advanced Transformer Techniques → Position embeddings (RoPE, ALiBi, T5 bias) → Layer normalization and sparse attention → BERT deep dive and finetuning → Extensions of BERT ➡️ Lecture 3: LLMs & Inference Optimization → Mixture of Experts (MoE) explained → Decoding strategies (greedy, beam search, sampling) → Prompting and in-context learning → Chain-of-thought reasoning → Inference optimizations (KV cache, PagedAttention) ➡️ Lecture 4: LLM Training & Fine-tuning → Pretraining and scaling laws (Chinchilla law) → Training optimizations (ZeRO, model parallelism) → Flash Attention for 3x speedup → Quantization and mixed precision → Parameter-efficient finetuning (LoRA, QLoRA) ➡️ Lecture 5: LLM Tuning → Preference tuning → RLHF overview → Reward modeling → RL approaches (PPO and variants) → DPO ➡️ Lecture 6: LLM Reasoning → Reasoning models → RL for reasoning → GRPO → Scaling ➡️ Lecture 7: Agentic LLMs → Retrieval-augmented generation → Advanced RAG techniques → Function calling → Agents → ReAct framework ➡️ Lecture 8: LLM Evaluation → LLM-as-a-judge overview →Best practices and benefits →Biases and pitfalls ➡️ Lecture 9: Recap & Trending topics From Stanford Online: Rigorous instruction. Latest techniques. Free access. Perfect for: → ML engineers building with LLMs → AI engineers understanding transformers → Researchers working on language models → Anyone learning beyond API calls This weekend: learn the techniques that separate good engineers from great ones. (I will put the playlist in the comments.) ♻️ Repost to save someone $$$ and a lot of confusion. ✔️ Follow @techNmak for more AI/ML insights.

@tedx_ai I've been trying to get BrowserTools to work in Cursor for several hours with no luck. Windows 11. I have a different MCP server (not yours) running successfully. It runs with npx @agentdeskai/browser-tools-mcp@1.1.0, but Cursor can't see it. How do I debug?

Wow the stability of @SlackHQ has gone down drastically in the last few months. Huddles are basically unusable (audio/video quality), and basic features like tagging and rich snippets are no longer reliable. Back to Google Meet for web meetings.

@SquirrelNebula @seanfrank Do you have your own ecom store (Shopify, Bigcommerce, etc)?

@seanfrank My company just started using amazon. We don't have to deal with this yet, but we get returns that are just empty boxes and there is nothing we can do. We sell record players that come back destroyed, nothing we can do. It's just lost money... But it's where everyone goes. 🤷‍♀️

Amazon is a broken mess. I have done over $50,000,000 on amazon. They dont care about me as a seller, partner, or person. I uploaded this design patent Was told “stop contacting me” What went wrong inside amazon? why do they hate small businesses and the law? A thread:

@asanwal They have a bad combination of 1. A superiority complex 2. No actual career accomplishments to speak of

"At HBS...." "When I was at Stanford..." If someone finds a way to drop this in the first 3 sentences of every conversation and is 30+ yrs old, what conclusion, if any, should be drawn?

The Supreme Court's Chevron ruling may be most impactful things to happen to startups in a long time, in ways that people don't realize. A thread:

This happens all the time. Universities very literally are not training you for a job as a software engineer. They are training you in the academic field of the science of computing.

Artificial intelligence and automation are already causing friction in the workforce. Should schools revamp existing programs for topics like #AI, or are new research areas required? @CIOdive @its_ahickey #artificialintelligence #highereducation #college replug.clouddatasummit.com/7f0f9420

Join the Ultimate Growth Hacking SaaS Giveaway and Win $110,000 in Tools and a Ticket to the Growth Marketing Conference in San Francisco. Thanks @startupbill dealify.io/giveaway?refer…

Microsoft leans on Azure, AI as cloud business grows #cio @CIOdive @industrydive replug.clouddatasummit.com/89ac9a60

Microsoft pushes AI as GitHub Copilot for Business hits 27K customers #cio @CIOdive @industrydive replug.clouddatasummit.com/847ec0a0