SWE-1.5 is a frontier-sized model with hundreds of billions of parameters optimized for software engineering. It achieves near state-of-the-art coding performance and sets a new standard for speed. The model is now available in Windsurf, serving at up to 950 tokens per second. It can be used to deeply explore and understand large codebases, build end-to-end full-stack apps, and easily edit configurations without needing to memorize field names.
2025.10.30·cognition.ai·8 min read·TLDR AI Agent labs ship product first and build infrastructure later. They turn AI models into goal-directed systems that deliver outcomes, capturing the real value in the AI stack. Founders only need a deep understanding of a domain and the ability to build reliable workflows on top of existing models. The most valuable developer skills are shifting from model architecture to system design, evaluation engineering, and domain-specific workflow optimization. Investors should look for companies that capture workflow data and have clear evaluation metrics. The moat is in the data and feedback loops, not the models.
2025.10.30·nibzard.com·8 min read·TLDR AI Mercor's marketplace connects former employees of investment banks, consulting houses, and law firms with the AI labs looking to automate those industries. Its customers include OpenAI, Anthropic, and Meta. The startup pays industry experts up to $200 an hour to fill out forms and write reports for AI training. It has tens of thousands of contractors that it pays more than $1.5 million to daily. Mercor's annual recurring revenue has grown to roughly $500 million in just under three years.
2025.10.30·techcrunch.com·5 min read·TLDR AI Sponsored
Spinning up AI agents has never been easier, but 62% of teams lose time or revenue to reliability issues. What's the missing piece? Read
2025.10.30·temporal.io·Sponsor·TLDR AI A look at what happens behind the scenes when a new model drops.
2025.10.30·threadreaderapp.com·2 min read·TLDR AI Anthropic's Claude now runs on AWS' distributed AI cluster, which spans 3 states with a total of 1 gigawatt of power.
2025.10.30·aboutamazon.com·5 min read·TLDR AI Saddled with lawsuits over teen suicides and heightened legal scrutiny, the AI roleplay startup will start phasing out teen chat access by November 25.
2025.10.30·blog.character.ai·2 min read·TLDR AI Scale AI and the Center for AI Safety tested frontier models on 240 real freelance projects sourced directly from platforms like Upwork.
2025.10.30·remotelabor.ai·24 min read·TLDR AI Sponsored
In Metronome's upcoming webinar [4], Martin Casado, General Partner at Andreessen Horowitz, reveals how leading startups and public companies alike are building monetization operating models for the AI era. You'll learn:
2025.10.29·metronome.com·Sponsor·TLDR AI OpenAI has restructured into a non-profit with a controlling stake in its for-profit business. The nonprofit is called the OpenAI Foundation. It holds a 26% stake in the for-profit arm. The for-profit arm is now a public benefit corporation called OpenAI Group PBC. Microsoft holds an investment in the PBC that is roughly 27% of the company. Microsoft can now pursue AGI independently or in collaboration with third parties, and OpenAI can now jointly develop some products with third parties.
2025.10.29·cnbc.com·4 min read·TLDR AI PayPal's digital wallet will be embedded into ChatGPT so users can pay for items through the chatbot starting next year. The deal will also allow merchants to sell items on the platform. The move is an early part of OpenAI's efforts to broaden ChatGPT's use for e-commerce. The company believes its 700 million-plus weekly users will use the service to find items, similar to a human personal shopper. PayPal is attempting to position itself as a payments backbone for the upcoming era of agentic AI shopping.
2025.10.29·cnbc.com·3 min read·TLDR AI Substrate, a San Francisco-based startup founded by a protege of investor Peter Thiel, claims it has reinvented a key link in the microchip production process. It plans to use its innovation to build manufacturing facilities in the US. The startup has targeted a three-year timeframe for producing its first chips. Industry analysts say the high costs and enormous technical challenges in lithography give ASML, the sole global manufacturer of advanced lithography machines, a 10-year head start over any would-be competitors.
2025.10.29·links.tldrnewsletter.com·8 min read·TLDR AI India is OpenAI's second-largest market with 29 million downloads in 90 days, but accounts for only $3.6 million in revenue. Perplexity and Google have also deployed aggressive free-tier strategies to capture the country's 700 million smartphone users to generate advertising revenue.
2025.10.29·techcrunch.com·3 min read·TLDR AI Meta's capital expenditure ratio hit 37% of revenue in Q3 2025, the highest capex-to-revenue ratio in the company's history. It announced $75.5 billion in infrastructure deals between September and October. Meta is financing the infrastructure for superintelligent AI systems. Meta can't compete with OpenAI on model quality alone - it needs a lot of dedicated compute, exclusive access to the latest Nvidia chips, and a vertically integrated stack to iterate faster, train bigger models, and serve more users. If AI doesn't deliver a return on Meta's investment, Meta will still have to pay rent for the next 16 years.
2025.10.29·allenarch.dev·13 min read·TLDR AI A forthcoming Nvidia chip would level the playing field between Chinese and US AI labs, which Trump may use as leverage over Xi in wide-ranging trade talks. The move would shrink America's massive AI compute advantage dramatically, especially since China's domestic alternatives are still far behind the US. Export restrictions are the only way to maintain the compute gap.
2025.10.29·threadreaderapp.com·3 min read·TLDR AI Sponsored
Last week, Baseten announced 650 TPS and 0.11 TTFT on gpt-oss [11], the most performant API on Nvidia hardware. Get near limitless scale and 99.99% uptime on one of the best open source alternatives to GPT-5.
2025.10.29·baseten.co·Sponsor·TLDR AI Agent HQ transforms GitHub into an open ecosystem that unites every agent on a single platform. Coding agents from Anthropic, OpenAI, Google, Cognition, xAI, and other providers will become available directly within GitHub as part of the GitHub Copilot subscription over the coming months. Agent HQ's power comes from a unified command center called 'mission control' that allows users to choose from a fleet of agents, assign them work in parallel, and track their progress from any device, wherever they work. It also provides new branch controls, identity features, one-click merge conflict resolution, and new integrations for Slack and Linear.
2025.10.29·github.blog·10 min read·TLDR AI FARMER integrates normalizing flows and autoregressive models to estimate likelihoods and synthesize images directly from pixels. It uses a novel dimension reduction and distillation strategy to improve efficiency while maintaining fidelity.
2025.10.29·arxiv.org·19 min read·TLDR AI OpenAI is tracking toward achieving an intern-level research assistant by September 2026 and a fully automated 'legitimate AI researcher' by 2028. The AI researcher will be a system capable of autonomously delivering on larger research projects. OpenAI is betting on continued algorithmic innovation and dramatically scaling test time compute to achieve its goal of creating superintelligence within the next decade. The company's recent restructure will help support its aggressive timeline for AI research assistants while maintaining a commitment to responsible AI development.
2025.10.29·techcrunch.com·3 min read·TLDR AI Eli Lilly and Nvidia have teamed up to build a supercomputer to boost the discovery of new medicines. The supercomputer and its AI capabilities will be able to identify new molecules to speed up the typical years-long development timelines. It will also help improve clinical trials, manufacturing processes, and sales. Financial terms of the partnership have not been disclosed. The system is expected to be online by January.
2025.10.29·links.tldrnewsletter.com·4 min read·TLDR AI Perplexity has outlined its practical vision for workplace AI, emphasizing usefulness over hype. With features like Comet, email and voice assistants, and app connectors, the platform helps reduce distractions, scale user impact, and streamline work toward meaningful outcomes.
2025.10.29·links.tldrnewsletter.com·2 min read·TLDR AI Sponsored
Ragie is a fully managed RAG pipeline with multimodal support, structured chunking, and multi-index retrieval, delivering the right
Google, through Broadcom, has raised its planned monthly TSMC CoWoS wafer starts for next year.
2025.10.29·threadreaderapp.com·1 min read·TLDR AI Arvid Lunnemark worked at Stripe before co-founding Cursor in 2022.
2025.10.29·threadreaderapp.com·1 min read·TLDR AI Shield AI will unveil its AI-piloted X-Bat fighter drone, which features vertical takeoff, a 2,000-mile range, and AI piloting via its Hivemind software.
2025.10.29·cnbc.com·5 min read·TLDR AI The 2.9% investment makes Nvidia Nokia's second-largest shareholder.
2025.10.29·links.tldrnewsletter.com·2 min read·TLDR AI Sponsored
Think the newest LLM is the best for coding? Despite their shared strengths and flaws, each LLM has a unique and inherent style—a measurable “coding personality” that drives their distinct results.
2025.10.28·sonarsource.com·Sponsor·TLDR AI Anthropic is continuing its push into financial services by launching Claude for Excel, which lets users analyze and modify spreadsheets by chatting with Claude in a sidebar. It is currently in beta behind a waitlist.
2025.10.28·anthropic.com·4 min read·TLDR AI Mercor, a human resources and recruiting startup, has finalized a new funding deal that values the company at $10 billion. The company manages 30,000 contractors around the world who help AI chatbots learn how to think and speak like humans. Venture capital firms are eager to offer funding to founders at sky-high valuations because they fear missing out on the next AI tech giant. Mercor is currently being sued by Scale for allegedly stealing trade secrets.
2025.10.28·links.tldrnewsletter.com·5 min read·TLDR AI RL environments can be surprisingly complex and fun to create. This post explains what RL environments are, introduces the 'verifiers' framework, and also walks readers through how to create an environment for a benchmark called AgentDojo. RL environments are scenarios that LLMs operate in for evaluation or training. Designing them means essentially defining the maze, the rewards, and how the LLM navigates through it.
2025.10.28·sidb.in·22 min read·TLDR AI Google recently published an announcement that its Gemma model had discovered a new potential cancer pathway. It's important to note that the model didn't invent or theorize a method to discover potential drug candidates - the entire construction of the experiment was imagined and executed by humans. The model was used to probabilistically narrow a set of potential candidates to a manageable set that could be inspected by human review. Humans still possess a reasoning capability that is not present in any current AI system.
2025.10.28·mindprison.cc·5 min read·TLDR AI Sponsored
We love this practical guide from You.com [9] that walks through a proven framework to identify, prioritize, and document high-value AI opportunities. Learn how AI can drive the most impact in your organization, both internally and externally. Get the AI Use Case Discovery Guide [9].
2025.10.28·about.you.com·Sponsor·TLDR AI TOON is a compact, human-readable format designed for LLM input that allows developers to pass structured data to LLMs with significantly reduced token usage. It excels at uniform complex objects. Standard JSON is verbose and token-expensive - TOON conveys the same information with fewer tokens. In benchmark tests, TOON achieves higher accuracy compared to JSON while using significantly fewer tokens.
2025.10.28·github.com·GitHub repository·TLDR AI Thinking Machines Lab demonstrated that training smaller AI models by having them learn from their own mistakes, by grading them with a larger teacher model, achieves the same reasoning performance as RL at 9-30x lower cost. "On-policy distillation" combines the relevance of learning from your own outputs with the dense feedback of traditional distillation, reaching 70% on AIME'24 math problems in just 150 training steps compared to 17,920 GPU hours of RL.
2025.10.28·thinkingmachines.ai·25 min read·TLDR AI A new sampling method enables base models to reach single-shot reasoning performance comparable to reinforcement learning, without retraining or external verifiers, while maintaining diversity and multi-shot accuracy.
2025.10.28·aakaran.github.io·6 min read·TLDR AI The smartphone chipmaker's new AI200 and AI250 inference accelerators—shipping in 2026 and 2027—sent shares up 20% on the promise of diversification beyond a stagnating mobile market. Saudi Arabia's Humain is committed to purchasing roughly $2 billion worth of the chips starting next year.
2025.10.28·links.tldrnewsletter.com·2 min read·TLDR AI Chinese Lab MiniMax open-sourced M2, a 230B parameter model with 10B active that ranks #1 among open-source models on Artificial Analysis' composite intelligence score. It slightly outperforms the previous generation Claude Sonnet model on coding benchmarks at less than 10% of the cost.
2025.10.28·minimax.io·3 min read·TLDR AI Sponsored
Access frontier models using the OpenAI SDK with integrated Snowflake tools. Build secure, enterprise-ready AI apps with state of the art inference and built-in control over data privacy and compliance. Get
2025.10.28·signup.snowflake.com·Sponsor·TLDR AI 2025.10.28·signup.snowflake.com·TLDR AI ChatGPT's mobile app download growth and daily active use are slowing.
2025.10.28·techcrunch.com·3 min read·TLDR AI AI search engines tend to cite less popular websites - ones that wouldn't even appear in the Top 100 links listed in an 'organic' Google search.
2025.10.28·arstechnica.com·4 min read·TLDR AI Following recent safety controversies, OpenAI revealed that approximately 1-3 million users per week show signs of mental health emergencies, suicidal ideation, or emotional reliance on ChatGPT.
2025.10.28·links.tldrnewsletter.com·10 min read·TLDR AI Sponsored
Pressured to use AI, but not sure how? Miro's put together a masterclass series to share real-world examples [4] of AI helping product teams.
The second installment is contingent on OpenAI restructuring to a for-profit by the end of the year, paving the way for an eventual IPO.
2025.10.27·links.tldrnewsletter.com·2 min read·TLDR AI