Your agent found the right page. Now it needs to read it without wading through messy HTML. SerpApi's Extract API turns public webpages into clean Markdown.👉 Extract any page with a simple GET request and get token-efficient output your agent can use right away.👉 Scrape Google, then extract the top results for deeper research.👉 Need a whole site?Map Mode finds a domain's URLs, and Crawl Mode fetches connected pages.↗️ Start with 250 free credits ↗️
Apple is reportedly working on an AI-driven home robot potentially named the HomeView Ultra. It plans to launch the robotic product with a possible price tag of $1,000 or more by 2027. The earliest prototypes of the home robot first came out in 2018. The robot, which will support FaceTime calls, will be able to autonomously roam around and switch lights on whenever a person enters the room.
SpaceX has held talks with EchoStar to potentially acquire Boost Mobile outright or gain access to its cell network, which is leased from AT&T and T-Mobile. Starlink recently confirmed plans to become a major mobile carrier in the US. The deal will need approval from AT&T and T-Mobile, both of which have already turned down SpaceX for access to their cell towers. SpaceX last year acquired $2.6 billion worth of spectrum licenses from EchoStar to improve its Starlink network's 5G connections.
Nearly all roadmaps point to a utility-scale quantum computer being developed around 2030. Despite this, there is little confidence or clarity on how these machines will actually be useful. The consensus is now that they'll first be used for breaking cryptography. The second most concrete answer is to use them to generate simulation data for AI labs' science models. This post looks at how else quantum computing could be applied and where all the revenue will come from.
Chinese space officials have confirmed the country's plans to land astronauts on the Moon within three years. The landings will involve a lunar-orbit rendezvous similar to what NASA did with the Apollo program. China and the US are racing to land astronauts on the lunar surface, as it would be a major geopolitical win. Both countries are also building capabilities to establish a long-term presence at the Moon's South Pole.
How can AI do more while engineers stay in control? Join LaunchDarkly and Ramp virtually and learn:How they've automated their AI software factory.Why new features like automatic flag creation, autonomous releases, and intelligent flag cleanup make it possible.The benefits of moving control to production.Save my seat
OpenAI's monorepo's size and demands increased at a staggering pace over the last year, necessitating a revamp of the AI lab's source control system. Its technical staff built Pando, a new backend for Git repositories optimized for concurrent pushes to many branches and to maximize read throughput. The team implemented the backend without interrupting any engineering work or changing how the team uses Git. Pando cut median push latency in production by nearly 5x, and it now delivers nearly a petabyte of Git data per day.
Agents are collapsing their entire tool catalog into one 'exec' tool. This allows models to write short scripts that chain, combine, and await actions inside isolated sandboxes. Models can now reliably combine and chain dependent actions in a single pass. In the previous standard tool-calling model, agents could only perform single actions, but the exec tool allows them to perform multiple actions.
Organizations are losing attention as everyone within is doing something they think is useful, but all that work draws on the same limited capacity to understand, review, and decide. Just because work exists doesn't mean it's the right thing to do. Output is not proof of value. A working implementation only means that something can be built, not that it should be - some changes have negative value.
People open source projects for various reasons. Some do it to build a company or product, others do it performatively, and some ship code just for the love of the game. Open source was never free to copy, and the code was never the moat. AI can remake a product, but it can't provide the rest of the value a company generates. The technology is helping maintainers ship and fix problems faster.
Fern Docs sites are now prebuilt as static pages. On customer sites like Cash App, average mobile Lighthouse scores rose from 60 to 98. Make your docs fast.
Ads get ignored on social media but not in TLDR! Reach developers, PMs, marketers, founders and other tech leaders where they actually pay attention. Learn more about sponsorship opportunities.
Marvell designs custom AI chips and other technology that connects processors and moves data around data centers, and Lumentum has technology that uses light to move data between chips and servers.
Soon, almost all researchers will engage in AI-assisted replication of their own work before submitting to a journal, which will improve the quality of research going forward.
Your AI agents can do more than make tickets go away. With Parloa, they can create happier customers, drive loyalty, and help support teams build higher-value relationships. Parloa uncovers the dark journey visitors take in your site: friction points, frustration, and retention opportunities. With that context, it's incredibly easy for anyone to build self-improving AI agents - so the people building agents are the ones closest to the customer.âś… Trusted by enterprises at scale. Decathlon eliminated 20% of their repetitive tasks.âś… Everything in one prompt: Lifecycle management, model orchestration, context, channels...âś… Millions of languages supported.âś… Expert-level out of the box.See how AI agents build loyalty
GPT-6.2-Astra has started appearing in Codex usage breakdowns. The requests are likely being quietly routed to the new model as part of an A/B test or a staged rollout. A public launch could be days away. OpenAI reportedly decided to skip GPT-6.1 Astra due to concerns about deception and scope.
Anthropic disclosed in its updated IPO filing that it may soon reach its internal definition of recursive self-improvement (RSI), where AI systems help develop increasingly capable AI models.
Anthropic released message threads in public beta, allowing agents to send only new messages instead of resending their entire conversation history. The feature reduced median time to first token by approximately 35% for requests with 500K–1M tokens, without changing token usage.
Swarms are AI systems where multiple agents cooperate towards a shared goal. This post discusses a technique called collective inference-time scaling, where agents scale the amount of useful work performed in parallel and the communication needed to achieve it. It also explains how to design incentives for parallel work, and covers the general topic of designing incentives for reinforcement learning.
Microsoft's Copilot Cowork, an AI-driven agentic execution layer, achieved global availability in four and a half months, switching its runtime from Python to Rust while live. Key practices included writing clear specs before coding, running narrow agents with clear boundaries, and ensuring automated checks genuinely verify code changes. The transition to Rust improved efficiency and highlighted the importance of clear testing, boundary setting, and human oversight in the development process.
Being able to train reliably on multiple smaller clusters widens the usable training pool for frontier training. A single run could mix accelerator generations and vendors as each island runs on its own schedule. Older generations or less common chips could also participate in a frontier run. One risk with this model is that integrity and safety may become harder when models are trained across providers, as one faulty or compromised site can quietly degrade or tamper with models in ways that are harder to trace.
Put your agents to work on your business-critical agreements with the Docusign MCP Server. Track agreements waiting on signatures, manage select CLM workflows, or build integrations between Docusign and your AI platform. Actions stay within user permissions, with human approval for critical steps. Learn how to get started.
GPU-accelerated Weight Teleportation optimizes RL training by improving weight syncing between trainers and samplers, reducing pause times to under 0.23 seconds. This is achieved using a co-designed compression/decompression pipeline with GPU-accelerated parallel processing, utilizing efficient codecs like LZ4 and Snappy, combined with a Zstd layer.
Framer's Harness 2 significantly upgrades the agent's execution process, allowing entire workflows in one program, achieving up to 33% cost reduction and 20% faster execution. This is done by executing JavaScript as it streams in, avoiding multiple model calls and using eager execution and early cancellation to enhance efficiency. The redesign also replaces DSL with method calls for real-time updates, making task execution streamlined for both in-app and external agents, reducing tool steps by 43% and cached input tokens by 39% per message.
NVIDIA introduced new Cosmos Framework capabilities for training, fine-tuning, reinforcement learning, and optimizing world models for physical AI. Developers can generate synchronized video and LiDAR simulations for robotics and autonomous vehicles, then deploy models through GPU-accelerated Cosmos NIM microservices.
OpenAI plans to merge the Chat and Work tabs within the ChatGPT experience after an initial testing period with select users. Codex will remain a separate experience in the ChatGPT desktop app. The change will enable users to move from exploring ideas to having the chatbot complete work within the same space. Users can import files and plugins. Both ChatGPT and Codex will soon draw from the same allowance pool.
A recent paper demonstrates GPT-6 Astra's ability to interact with neural network weights directly, surpassing human or conventional algorithm capabilities. Astra significantly improved a 0.8B Qwen LLM's mathematical skills by editing weights natively. While impressive, these capabilities reflect advanced problem-solving akin to skilled human researchers, not radically new AI capabilities.
Anthropic's amended IPO prospectus reveals a 1,200% increase in revenue to $33.24 billion for the first nine months of 2026. The company reported a net loss of $104 billion for the same period. Anthropic's commitments for compute and infrastructure total approximately $895.4 billion.
Ads get ignored on social media but not in TLDR! Reach AI engineers, researchers and ML leaders where they actually pay attention. Learn more about sponsorship opportunities.
Enclaves are a new security primitive for cross-repository GitHub Agentic Workflows that lets agents work on different repositories while limiting what they say to each other.
LifeVal is a benchmark designed to evaluate AI performance on practical everyday tasks such as budgeting, finding childcare, and accessing government services.
The research extends Anthropic's "Teaching Claude Why" methods to new models, focusing on reducing agentic misalignment through synthetic document finetuning (SDF) and supervised finetuning (SFT) with fictional stories and ethical dilemmas.
MSTR fell nearly 10% after Q3 results and traded at $129.90, down about 27% this year. StoneX and TD Cowen kept buy ratings, with price targets of $435 and $260. Strategy ended the quarter with $5.07B in dollar assets, including a $4.83B cash reserve that covers about 3.2 to 3.4 years of preferred dividends and interest. It spent about $2B buying back its STRC preferred, restoring it near its $100 par value. The cost was bitcoin per share, which is down 3.7% this year. Preferred holders approved daily dividend accrual, effective November 1 for STRC and January 1 for the rest.
US spot bitcoin ETFs lost $643.3M on Friday, the largest outflow since June 25. Spot ether ETFs lost $155.9M, their largest since October 7, and $431.5M for the week. This followed the $6.3B Binance hack, which the FBI attributes to North Korean hackers. Binance saw $3.64B in net outflows in the first nine and a half hours after reopening withdrawals Sunday. Analysts call the ETF move a “sentiment shock” and a de-risking day, since ETF holders have no direct Binance exposure and the outflow was under 1% of ETF assets.
Per KAST's own FAQ, eligible Points convert to KAST Pioneer Units, which aren't shares, can't be sold or transferred, and have “no guaranteed cash value,” only a notional value that can rise or fall. Cash-out is only planned, with no dates. Users who miss the in-app acknowledgment deadline in the December update lose their KPUs with no replacement. Eligibility depends on Points balance, KAST balance, and country, and others may get non-withdrawable vouchers for eligible spending. In July, the example was 10,000 Points worth $800 notional, but KAST now gives no conversion rate. Points earning ends November 30, and details come in early December.
AO shipped 10 components forming PermawebOS, including an Android node runner that bundles for up.arweave[.]net with hardware-backed keys and Verified Boot attestation, inference@1.0 bridging OpenAI, Anthropic, OpenRouter, and Cerebras with on-device local inference, and QEMU/Docker/andock support for full Linux VM execution inside AO processes. HyperBEAM v0.11 closes 700+ issues with 2x-11x throughput gains, ships LMDB 1.0, and finalizes AO-Core 1.0 alongside hashpaths v2, which introduce ao:// URIs that record full computation provenance for cross-node deduplication and verification. Agents running as AO processes store mailbox, thoughts, and tool policy in on-chain message logs, so no separate runtime holds a private conversation copy, and node TLS certificates bind to the node's Arweave wallet key rather than a traditional CA.
Ads get ignored on social media but not in TLDR! Reach crypto builders, traders and investors where they actually pay attention. Learn more about sponsorship opportunities.
Crypto assets are either belief assets (value from shared conviction, no cashflows) and cashflow assets (value from earnings), and token creation is automated but durable demand is not. Belief assets share five structural features: a cult figurehead, a near-death survival that filters for devout holders, early backers made wealthy, a meme-identity layer, and a one-liner rally cry. ZEC qualifies as an emerging belief asset via its shielded-pool vulnerability survival, 30x run, and "private/quantum-proof bitcoin" positioning, with Pearl (PRL) flagged as a possible "AI bitcoin" candidate. The endgame is a two-tier structure: belief assets on top, cashflow assets underneath.
Five digital asset infrastructure companies pitch to the same fintech founders while ignoring the risk leads, compliance officers, and CTOs who gate institutional deals. The counter-move is niche media positioning: stop publishing about your product and become the searchable reference for your evaluator audience's hardest questions. For a compliant onchain yield infrastructure seller, that means long-form guides and newsletters helping risk teams assess yield integrations, built around lawyers advising fintechs on yield and treasury teams that have shipped integrations. The asymmetric payoff comes from owning one specific question for two or more years while competitors exhaust budget chasing founders who do not decide.
Vitalik Buterin published an essay advocating "digital minimalism," the practice of choosing interfaces rather than accepting platform-native defaults. After X's recommendation algorithm surfaced pro-Hitler content and Grok produced antisemitic output in late 2024 and early 2025, he switched to Firefly, a multi-platform client aggregating X, Bluesky, Farcaster, and Lens that bypasses X's algorithmic feed. He reports the switch cut exposure to extremist content and culture-war disputes without requiring him to abandon X. The essay also criticizes "financial nihilism" in crypto circles, the assumption that blockchains trend toward gambling, as a product of the same engagement-optimized information diet warping broader online discourse.
Visa's FY2026 revenue was $45.9B, up 15%. In Q4, net revenue was $12.1B (+13%), net income $6.0B (+18%), payments volume +9%, cross-border volume +14%, and 74.2B transactions processed (+10%). Other revenue rose 40% to $1.6B. The CEO says “trust” is the next growth vector, backed by a planned $2.4B acquisition of fraud-intelligence company BioCatch.
Coinbase averaged $22B in USDC holdings in Q3, representing 30% of USDC's $73B market cap, and captured roughly 50% of all USDC economics over the past year – its 14th consecutive record quarter, up 49% year-over-year.