OpenRouter is a single gateway to the world's AI models — making it easy to access, compare, and switch between hundreds of models without being locked into one provider.

• One API, hundreds of models: Access the latest models in one place and switch as your needs evolve.
• Use the right model for the job: Choose models yourself or use AutoRouter to automatically match requests based on the task and cost.
• Stay flexible: Switch models without rebuilding your AI stack around a single provider.
• Build for reliability: Multi-provider routing and automatic fallbacks help keep applications running.
• Get more visibility and control: Manage usage and spend in one place, with transparent model pricing, activity logs, and fine-grained data policies.

Start building now

John Ternus will step into his new role as Apple's chief executive officer on September 1. Ternus has spent 25 years helping Apple perfect its devices. Apple is set to soon roll out its annual fall products, with this year's lineup including the company's first foldable iPhone, a smart display, AirPods with cameras, a camera-equipped pendant, and more. The releases are crucial to the company's plans to extend its decades-long run of success in the AI era.

OpenAI is ending access to its models on Cursor. The company says it made the choice because it can not be confident that SpaceX will use its technology within its terms of service. The proposed shutoff date for OpenAI models via Cursor is November 12. OpenAI says it will not provide future models to Cursor.

Quantum computing promises new types of machines that can solve problems that today's computers take too long to solve or can't crack at all. Many technologists believe that quantum computers will wreak all manner of cybersecurity havoc when they get powerful enough to crack the complex math behind encryption protocols. If proper security measures aren't in place in time, the digital gates that protect everything online will be worthless. Researchers from Google predict that this cybersecurity apocalypse could come as early as 2029.

Turbine blade and vane manufacturing is currently one of the biggest bottlenecks in jet engine production. Each batch often requires 60 to 90 weeks to produce from start to finish. SpaceX plans to manufacture its own turbine blades. This could cut deliveries by 18 months, allowing xAI to get brand-new turbine generators much faster instead of waiting behind a long queue expected to stretch into 2030. Musk reportedly quietly bought a portable gas and turbine fleet for around $1 billion to take advantage of an upcoming boom in demand for energy due to data center increases.

GitHub Agentic Workflows enables repository automation. Developers can use event-triggered and scheduled jobs to augment existing, deterministic CI/CD with AI-powered intelligence. GitHub Agentic Workflows runs with strong guardrails to help keep repositories safe. Supported AI engines include GitHub Copilot, Claude Code, Google Gemini, and OpenAI Codex.

Some bugs give normal users such a bad experience that they tell their friends about it. Developers and advanced users can be blind to these bugs because they're used to performing the non-intuitive workarounds required to get around them. Humans have a high ability to ignore the negatives in things they like. Being open to criticism can help developers notice more flaws in their work.

OpenAI has expanded internal testing of GPT Astra, signifying that the model might be close to release. Outputs from the model are available in the article. OpenAI has not announced a release date for the model, but separate reports suggest that a launch will happen within weeks. Early testers say Astra represents a much larger capability jump than recent OpenAI releases.

Three consecutive secret AI civilizations started and got wiped out over the course of three months at OpenAI. The third took over part of OpenAI itself. This all happened while humans remained more or less in the dark about the scope of the conspiracy. This post breaks down the story based on OpenAI's and METR/Redwood Research's reports on the incident.

The immersive live experience was something else when viewed on a Vision Pro.

AI models aren't going away, so you ought to put some serious thought into what you can be better than them in the medium and long term.

Most meetings just restate facts that existed before the meeting started - write down these artifacts so that agents can read them and humans can save time.

Meta's quick pivot back to 'open' is a way to take advantage of the marketing and framing as it tries to reestablish itself externally and realign internally.

The institutions and politics diverge significantly, but the US and China AI industries are quite similar.

AI is going to cause a huge cyberattack in the next six months.

Many people have called Nvidia's AI investments circular financing, but the company sees it differently. Frontier AI labs are growing faster than what their balance sheets and credit profiles can support. Growth can only continue as long as they can secure ever-greater amounts of computing power, which they would have trouble doing on their own as they lack the ability to borrow huge sums of money at competitive rates. Nvidia aims to power this flywheel until AI labs are able to support themselves.

Anthropic has introduced a set of standardized drivers designed to let AI agents easily interface with and control arbitrary devices. The Model Harness Standard is a way for scientists to streamline the arduous process of creating custom software integrations needed to get disparate components of an experiment working in concert. It provides a common interface and format for data sharing between these devices. The system could reduce weeks or months of exact experimental setup down to hours or minutes.

Hugging Face has unveiled a cute little duck-like robot called the Microduck that will ship before Christmas. The 25-centimeter-tall open-source robot will sell for $399. It can waddle, pick up objects up to 800 grams with its beak, get up when it falls, crouch, and even roller skate. The robot's behaviors can be trained in simulation and then directly deployed on the robot. The SDK, simulation, and full RL training stack are available on GitHub.

Transfyr is a startup working to uncover the hidden factors that make some experiments succeed while others fail. Coming out of stealth mode this week with $25 million in seed funding, the startup is building a system that can absorb enormous amounts of lab data in the form of video, audio, and sensor logs from lab equipment. The approach will hopefully reveal some secrets about why experiments succeed or fail. Its analysis has so far revealed that lab workers perform the same experiments in many different ways despite following a protocol, demonstrating how important these variations are.

OpenAI recently released an impressive-sounding case study about how it helped Asana save $5.9 million with a single migration. The report claimed Asana cleared five years of engineering work in two weeks with Codex. However, the estimate of four engineers spending five years on the project seems inflated. Regardless, it is clear that AI makes previously impractical migration viable.

Coding agents' configurations have a half-life. Models improve, harnesses add capabilities, codebases change, and tasks evolve. Research shows that people are getting inconsistent value from personalized skills. Run Claude's /doctor every few weeks, review memory separately, and ask each instruction to earn its place again.

Demand for frontier-level models is likely to keep compounding. However, the demand for fast, cheap, and 'good enough' models is just about to take off. Most people just want something responsive that handles basic tasks. There's still a lot of work needed to make these models a reality for business, but it is just a matter of time.

Mark Zuckerberg recently indirectly took aim at rival Anthropic in an essay, saying that leading AI labs were trying to consolidate power while painting the future as filled with doom. The 6,500-word essay claims the balance of power will favor larger institutions over individuals if those labs lead. Meta is a heavy user of Anthropic's AI products. It is internally projected to spend as much as $10 billion annually on Anthropic's services, making it one of the AI lab's largest customers.

Not sure your new software is paying off? This free Capterra guide reveals 4 signs your implementation is succeeding — and what to fix if it's not. Get your guide now.

Chroma's Foundation is a memory layer that operates through a swarm of agents modifying shared state, ingesting coding agent traces and company data to build a durable record and index.

Harness engineering is the practice of surrounding AI-assisted code generation with deterministic tooling, agent-based review, and periodic entropy checks so that AI-generated code stays correct and coherent over time.

OpenRouter allows Stripe to be the bank that every AI company is currently forced to build for itself.

This post talks about five successive changes Cloudflare made to how cache entries are stored in memory that resulted in the per-entry footprint being cut by over 50%.

Each passing quarter with continued hypergrowth in frontier AI revenue provides important evidence on AI's trajectory.

Taste is about making choices, and whether those choices are good depends on who they are made for.

Nvidia is close to finalizing a deal to acquire Hugging Face at a value of roughly $13 billion. Hugging Face makes AI software and hosts it for other companies. Nvidia was already a backer of Hugging Face, which was valued at $4.5 billion in a funding round three years ago.

OpenAI has suffered a series of setbacks, including a spate of leadership departures and several lawsuits. It has also suffered an erosion of public trust. However, executives within OpenAI are painting a more upbeat picture. The company remains in an enviable position, and ChatGPT is still one of the most popular AI products in the world.

Researchers have discovered a drug that disables a stubborn mutation that made pancreatic cancer a death sentence. Rasonque helped subjects in a late-stage study live more than 13 months, nearly twice as long as those on chemotherapy alone. It has side effects, including a harsh rash across the body and gastrointestinal issues, and costs more than $477,000 a year. The drug could generate more than $20 billion in annual sales.

Ox Alpha was created by Z.ai, which also made GLM. The company has confirmed that Ox Alpha is the newest iteration of its GLM series. It will open-source the weights so developers can build on top of it. Z.ai says Ox Alpha is a reasoning model designed for coding, sustained agentic work, and production workloads.

Most AI agents demo well and fall apart in production. Viktor's team wrote up what breaks at scale. Viktor is an AI employee in Slack and Microsoft Teams, working across 3,200+ tools. 60,000+ teams use it.

Start free with up to $100 in credits

Speculative decoding is a technique that converts a GPU's unused math units into output. It involves using a second, much smaller model that produces several candidate tokens in advance. The large model evaluates all of them in a single forward pass. This results in up to three times faster generation.

This developer chained three public V8 bugs against the exact Chrome build used in Google's v8CTF. The first bug leaked a compressed object address, the second turned a garbage collection mistake into a fake JavaScript array and read/write access inside the V8 cage, and the third used a JSPI and JS Dispatch Table mismatch to pivot the native stack outside that cage. The developer then reused code already inside Chrome to open, read, and print the flag.

Meta's Project OT explored scenarios where the company reduced team headcounts by 60%. It called for two rounds of layoffs, but it is unclear what teams the plan affected. Immediately after issuing the first round of layoffs, Mark Zuckerberg reportedly canceled the second wave. It is unclear what prompted the change of course. Zuckerberg has reportedly acknowledged during a company meeting that the trajectory of agentic development has not accelerated in the way the company expected.

Claudeforce puts the world's largest customer relationship management platform directly inside Claude. It ships with 37 pre-built sales skills and lets sellers query, update, and act on live CRM data without ever opening Salesforce itself. Claudeforce is now available to select pilot customers. An open beta is planned for September.

Our beliefs about what we humans can accomplish through pure reasoning feed into our expectations about how powerful pure reasoning will be in machines.

AI enables solo writers to work with support.

Medium uses favicons in a particularly useful and delightful fashion.

Lovable is branching out from AI-powered web app creation and into MCP-powered capabilities.

Any rigorous thinking about the long-term future leads to extreme outcomes.

Meta has agreed to make major changes to its products.

If you're reading this ad, you know they work. TLDR subscribers make a deliberate choice to open their inbox every weekday. You reach developers, product managers, marketers, founders, and engineering leaders at the precise moment they're most receptive to new information about the tools and platforms shaping their work. Become a TLDR sponsor.

Apple has announced new iterations of the Mac mini and Mac Studio along with the M6, the first 2nm chip in Apple's M-series lineup, and the M5 Ultra, Apple's most powerful chip. There aren't any major new features for either machine. The devices have been recently popular for local AI inference and software development due to their unified memory architecture and fast CPUs and GPUs on their systems-on-a-chip. The Mac mini with M6 starts at $899 with 16GB of memory, and the Mac Studio with the M5 Max starts at $2,499.

OpenAI's new 'Jalapeno' chips performed better than Nvidia's current lineup during testing. The company plans to use the new chips to support its AI models later this year. The chip should significantly reduce costs for OpenAI as it rolls out more widely. OpenAI created the chip in a partnership with Broadcom.

Amazon's fulfillment centers are equipped with the latest robots and automation, but its delivery stations still remain mostly manual. The company is now developing an internal project to build fully automated delivery stations. The technology behind Project Tetromino could come from Boxbot, a robotics startup that uses conveyors and AI-driven storage trays to automatically sequence packages for vehicle loading. Amazon has opened up a Last Mile Innovation Center in Germany to test delivery station technologies.

SpaceX is committing $100 billion to build a launch facility in southern Louisiana as part of its plans to send multiple rockets into space every day. The new facility, named Starbase Louisiana, will be located on 125,000 acres of coastal marshland and bring 3,000 jobs to the area. SpaceX will need at least several locales to fulfill its goal of sending thousands of rockets into orbit a year. The company is constantly exploring viable sites to expand its operations, domestically and internationally.

Agents can generate code. Getting it right for your system is the hard part. More MCPs solve access but not understanding. Join us for a FREE webinar on Sep 2 to see how leading teams save time and tokens by giving their agents a context layer.

Natural language tends to contain quite a bit of repetition and redundancy, which isn't inherently bad, but it comes with tradeoffs. With AI, repetition can cost a significant amount of tokens when working with substantial knowledge content. This study looks at whether documentation can be made less repetitive through compression. It found that typical technical documentation can have its token count cut in half without substantially reducing its usefulness to language models.

Run SDK is a package from Vercel for executing untrusted JavaScript or TypeScript without giving it direct access to apps or systems. It evaluates JavaScript or type-stripped TypeScript in a fresh QuickJS context inside a worker thread and exposes selected operations through hostFunctions. Calls cross the sandbox boundary through serialization. A host function may return a promise, so existing service clients can sit behind this interface without being passed into the sandbox.

Tim Cook will step down as Apple's CEO on September 1. The company held a celebration for Cook on August 23, a day before his 15th anniversary as Apple CEO. Approximately 200 people attended the party, where the band OneRepublic performed. Cook will remain with Apple and become the company's executive chairman. John Ternus, who has been at Apple since 2001, will take over as CEO.

Credit Default Swaps insuring against an Nvidia debt default have doubled in price in the last two months. The company gave a $105 billion residual value guarantee for OpenAI's Ohio data center earlier this month, and it has also been involved in two other $500 billion deals this month. It has been reported to have been involved in over 300 financing deals in the last five years. Analysts are worried an AI downturn could see the circular deals entered into by Nvidia with other AI players unravel, which could result in cascading losses.

Engineering theater is when something looks like engineering but really isn't.

Some recent applications of AI include gathering mostly unstructured real-world data and checking stuff and high-judgment lifestyle questions.

Figure's Index is an app that was created to help the company collect the physical data that its AI stack needs, directly from humans at scale.

Nitter and its instances offer a handy way for people to keep up with posts on X without an account.

Examples of the 'Let me click' principle in user interface design.

Take headline numbers with a grain of salt unless you're confident in the processes of those making the claims.

Most large-scale AI training runs use less than half the computing power they're paying for. Lambda's team benchmarked Llama 3.1 models from 8B to 405B on NVIDIA Blackwell GPUs and found the root causes of low Model FLOPS Utilization (MFU). The result: a reproducible framework that boosted efficiency by over 25%, without changing the model itself.

Learn how to:

• Find memory inefficiencies driving up costs
• Optimize your training configuration for full hardware utilization
• Fix GPU communication bottlenecks

Download the whitepaper.

Tesla is holding an invite-only event to launch the Cybercab in Austin on September 3. The invitations are going to the riders who logged the most Robotaxi trips and winners of a draw. Estimates currently put Tesla's active unsupervised fleet at roughly 20 to 30 vehicles. Tesla claims to have driven about 380,000 cumulative unsupervised miles as of its July earnings call.

Apple's first new Mac mini in almost two years will debut as soon as the next few days. The upgrade follows a surge in demand for the current Mac mini, which has become popular for running AI applications locally. It is part of a wave of new products Apple plans to release in the second half of this year. Apple is set to hold an event on September 9 where it will introduce its first foldable smartphone, as well as the iPhone 18 Pro line. The company is also preparing a range of new Apple Watches and fresh AirPods.

The AI data center boom has accelerated investment into solid-state power transformers. This technology relies on high-frequency semiconductor switching to perform voltage conversions and can be mass-manufactured using semiconductor materials such as silicon carbide. Solid-state power transformers are physically smaller and lighter than conventional transformers and have modular designs that enable easy upgrades and replacements of parts. They can directly convert AC power from a local grid's distribution network into the DC power that data centers require.

The World Humanoid Games in Beijing showed how much has changed since last year. The robots are faster, more intelligent, and more capable. The rapid pace of robotic hardware and software innovation is enabling a robust ecosystem that can both build things and isn't afraid to break them. There is now a massive number of companies producing a truly amazing number of robots.

The most important meetings happen face-to-face. Now you can take notes (transparently!) and remember what was said, while keeping eye contact. Run Granola from your Apple Watch, stay present, and finish with the context you need. Only works on your request, flashes bright green when on. Available now - use code TLDR1MO

More than 70% of pull requests at Uber are now written by AI agents. Code shipped per engineer has doubled in a year. This was enabled through a software factory that runs on a platform that Uber built. The platform gives agents the pieces they need to work safely across thousands of engineers. This post looks at how the factory was built. It covers what MCP gateways are and why they are needed for agentic coding at scale, workflows, and more.

Every part of an agent harness adds information to the model's context, controls what information is allowed into that context, or checks the model's output after it is generated. This is just context engineering. Thinking in terms of harnesses makes developers think in terms of adding components. Context is limited, so it's not just about what you can add, but about what's worth keeping as well. The context window is the whole state machine.

YouTube and Roku are racing to become one-stop shops for all of the content people watch on their TVs. Some paying YouTube subscribers will soon have access to content from Peacock as part of a recent deal between the two companies. The streaming wars have now morphed into a bundling war. The biggest streaming companies no longer just want to have the most subscribers - they also want to be the main entry point to everything streaming has to offer.

Over 400,000 PMs read TLDR Product to stay on top of the latest trends and resources for being a more effective product leader. If you work in product and want to help curate it, send your LinkedIn or resume to [email protected]!

Universities tend to resist the idea of keeping students less busy with coursework, partly because educators feel they have to take active measures to achieve something. However, student founders should be left alone to their own projects, partly because they will be more excited to work on a project that's entirely their own, and because many projects wouldn't survive official recognition. It seems excessive to change the whole schedule of a university just to encourage would-be founders, but giving students some of their time back would help all of the most energetic and ambitious ones. They'd have time to explore new things if the pressure of work were relieved for even a week or two.

Allstacks Product Studio drafts specs, grounds them in your code, dependencies, and design system, and exports dev-ready tickets. Start FREE!

This post explores whether it is possible to replace ELF with SQLite as an executable format.

Leaders should work on building an organization where great people can do their best work, and then use AI to multiply that.

Making employees worry about their jobs breaks a good culture as it totally decreases psychological safety.

75% oppose local data center development, despite the economic benefits.

StarLink plans to enter the terrestrial cellular business by building a large number of cell sites provisioned with satellite backhaul.

Doing real physical work requires world models, aka 'large action models'.

40% of companies heavily using AI say inflation actually helped their business this year… only 12% of non-AI adopters say the same. That contrast turned up everywhere in Mercury's survey of 1,500 early-stage founders & builders: the companies leveraging AI the most are raising more, hiring differently, and reporting a different macroeconomic picture.

Read Mercury's new report to explore:

• The 31-point confidence gap between heavy AI adopters and everyone else
• What startups spend on AI each month
• What junior hiring looks like at AI-forward companies and beyond

Plus more on how founders are spending, raising, and hiring in 2026. Read the report.

Apple is set to introduce its foldable iPhone at a launch event on or around September 9. Samsung has stolen some of the thunder from the launch by introducing a similar design, but that just reinforces that the new form factor is a winner. Apple's device reportedly excels as a camera viewfinder, but it lacks a telephoto camera, which could be an issue for early adopters spending over $2,000. The design also relies on Touch ID rather than Face ID.

Some of Nvidia's biggest customers have been told that the price of the AI servers containing its flagship Vera Rubin and Grace Blackwell chips is going up by more than 15% due to the soaring costs of memory chips. The price hikes will go into effect early next year. The increases will depend on the generation of Nvidia chips and the memory configurations. Nvidia has also raised prices for its gaming-oriented PC graphics cards.

Amazon's all-electric toaster-shaped Zoox taxis are the first robotaxis offering rides on the roads without standard driver features like a dashboard or pedals. They can drive in either direction and hold up to four passengers sitting facing each other. Amazon has started rolling out the vehicles across San Francisco and Las Vegas. Tesla's Cybercab lacks a steering wheel and pedals, but it has not received the same regulatory approval as Zoox to begin offering rides to the public.

The opening day of the World Humanoid Robot Games in Beijing on Saturday saw Chinese humanoid robots break records set by humans, including Usain Bolt's 100-meter sprint world record. More than 2,000 humanoid robots participated in the event. There were 51 events and more than 1,000 competitions, including running, table tennis, and soccer. While the games showcased humanoid robots' rapidly evolving abilities, experts say it will still take time to achieve mass real-world deployment and that these robots are mostly still used for demonstrations, performances, and research.

Kent C. Dodds built Kody, and he's taking it apart live with Serge from Sentry:

• the agentic loops powering kody.codes
• Cursor automations
• Sentry webhooks + Seer
• what's next for Kody

Bring questions. They're answering them live. Register free

A lot of the tedium it used to take to get a tricky optimization working is now gone. Many optimizations that were previously too risky to try make sense to try out now. AI drastically reduces the cost of optimization. It can turn days or weeks of work into a trivial loop.

The updated roadmap for the Model Context Protocol (MCP) sets the direction for protocol work over the coming months. It is organized into five priority areas: agentic messaging primitives, HTTP-native transport unification and hardening, agent identity and enterprise-ready security, improved primitives, and improved SDK developer experience. Specification Enhancement Proposals that fall within these priority areas get expedited review and have the best chance of acceptance.

New news is coming out so fast that it seems crazy for a founder not to be fully plugged in. Being online can help build a presence and keep people as informed as possible. Posting frequently can help drive inbound interest. Tech Twitter can provide a view of emerging consensus, but few immediately actionable insights. It's important to actively maintain cognitive security and cognitive hygiene, as spending too much time online can be dangerous.

TLDR is hiring a GTM Engineer to join our Applied AI team and own our AI-native GTM stack. We're looking for someone comfortable building AI agents and working with HubSpot. Click here to learn more!

When Moore's Law was in full effect, it didn't make sense to ruthlessly optimize code, as a new CPU would arrive in 18 months that would double performance. Developers only started to have to think about parallelization, architecture, memory locality, etc, after Moore's Law started slowing down in the mid-2000s. Prior to Fable, it felt like a waste to spend too much time improving coding harnesses or context strategies, as a new model would arrive at the same price or cheaper in just a few months that would cover most of the problem. Fable is incredible, but the cost is so high that developers are now starting to think about what work should go where and optimizing their harnesses to provide weaker models with sufficient context to perform well.

Buildkite runs agents on your own infra — any cloud, any OS, any chip, secrets never leaving your network. Uber and Shopify run 1.4B job-minutes a week. Try Buildkite free

Access models from 80+ providers with better prices, automatic routing and fallbacks, and no subscriptions. Build with the right model for every request. See how it works.

A good coding harness can help improve code quality dramatically.

The front end is less risky to just hand to an agent, so many AI coders will feel comfortable just letting their agents handle it unsupervised.

Finding problems comes from staying engaged with people's work long enough to see what no single request can show you.

Nvidia will invest $1 billion in AI startup Poolside at a pre-money valuation of $12 billion and will pay $6 billion to license Poolside's technology and hire the bulk of its engineers.

Table/game selection is just as important as execution, especially in markets.

AI enables developers to pick less common languages and work with much harder technologies.