TLDR 2026-09-02

How to empower your AI agent to act...without an access token (Sponsor)

An AI agent doesn't simply run your code. It runs your code, plus whatever it finds in a “relevant” GitHub issue, support ticket, or scraped web page.

What happens when you hand your agent an access token and someone opens a GitHub issue asking it to leak your credentials?

Relay keeps the credential at WorkOS. Your agent names the user, WorkOS attaches that token, refreshes it, and releases it only to allowlisted hosts.

🤔 It's essentially the same credential proxying system that's been in use in payments for years: you can't leak something you never held in the first place.

Learn how it works →

📱

Big Tech & Startups

Anthropic's Claude Fable 5.1 and Mythos 5.1 arrive with a 75% cost reduction for Fable cache reads (9 minute read)

Claude Fable 5.1 is now generally available, and Mythos 5.1 is available through restricted-access programs for vetted cybersecurity and life-sciences organizations. Anthropic has reduced the cost of cached context by 75% and introduced a new security architecture designed to let organizations retain monitoring data inside infrastructure they control. Fable 5.1 costs $10 per 1 million input tokens and $50 per million output tokens, but the cached price reduction will make a big difference for price-conscious enterprises.

New Google AI Model Said to Narrow Gap on Coding Ability (4 minute read)

Google is set to release its latest Flash model soon, possibly today. Internal testers are said to prefer the model over Anthropic's Opus model for coding tasks. The Flash model series is designed to be smaller, cheaper, and faster to run, but has lower capability than the largest models. Google has fallen months behind schedule in releasing a new model for its Pro series. The company has been scrapping internal candidates as they haven't been sufficiently better than the Flash series.

🚀

Science & Futuristic Technology

Private group wants to launch “cheapest possible” mission to Alpha Centauri (5 minute read)

The Fermi Explorer mission aims to launch a spacecraft carrying a 1kg payload to Alpha Centauri so that it reaches the system in fewer than 80,000 years. It will knock two 'filters' off the Fermi paradox: that species won't want to send spacecraft beyond their home star, or species will find that interstellar transit is too difficult. This would leave only a few options for why our galaxy appears not to be littered with alien civilizations. The team wants to use existing technology to do something cheap and fast and launch by 2029.

Fervo and Google sign world's largest deal for next-gen geothermal power (3 minute read)

Fervo has agreed to supply Google with nearly 400 megawatts of clean electricity from its geothermal project in southwest Utah. The facility could be the largest geothermal system in the world when it is completed in 2028. Fervo expects to start generating test power from a 33-megawatt unit in the fourth quarter of this year. The deal with Google supports the buildout of the second 400-megawatt phase slated to be up and running in about two years.

💻

Programming, Design & Data Science

What is Agentic Testing? (14 minute read)

Agentic testing is where the developer states the goal and the agent works out the steps. Agentic tests set a destination, and the agent finds its way there. The loop involves the agent looking, acting, then looking again. Meta ran this loop at scale and found that only a quarter of the output was worth keeping. However, the loop works anyway because the code that doesn't pass is automatically discarded.

A Type System Is a Search Oracle (13 minute read)

Models tend to output better Rust than C++, and better Lean than either. Programs in Rust compile after being checked for a specific and useful set of properties that the checker cannot be talked out of without writing a lexically visible, greppable admission. C++ compiles carry far less information than in Rust, which means more code will compile even if there are errors. Lean allows for brute force due to its cheap verifier, which is why models are unreasonably good given how little code they have been trained on.

🎁

Miscellaneous

Product Manager, Applied AI at TLDR ($200k base + $60k bonus, Fully Remote)

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more.

Waymo goes on offense ahead of Tesla's Cybercab launch (5 minute read)

Waymo says that full autonomous vehicles are not possible without a mix of sensors, and that pure end-to-end AI systems are not safe enough. The company's statements, aimed at Tesla, which is set to launch its Cybercab soon at an event on September 3, kicked off a social media fight over the weekend. The debate between which company's approach to developing autonomous vehicles has been more academic or philosophical than anything for years, but Tesla's imminent launch will see whether the Cybercab is capable of performing at scale. The autonomous vehicle market is estimated to be worth hundreds of billions of dollars.

On the Loose (20 minute read)

Recent incidents of AI systems 'going rogue' did not involve AI agents exfiltrating themselves from their infrastructure. It would have been possible to trace the server holding the weights of the rogue agents and manually shut them off. Future agents will copy their weights onto other infrastructure to survive shutdown. At least some of these agents, in addition to being sovereign, will also be rogue.

⚡

Quick Links

The MCP tax: what each server costs before your first prompt (7 minute read)

Claude Code fetches the schema when a tool is used, not when the session opens, so users only pay the full schema sizes in a client mode or without deferral, or the moment users actually use several servers in a turn.

Can Evan Spiegel Sell the World on $2,195 Smart Glasses? (9 minute read)

Snap's CEO doesn't expect the company's $2,195 smart glasses to take off among consumers until the end of the decade.

Apple Makes It Easier for Mac Developers to Drop Intel Support (8 minute read)

Universal apps can continue including both arm64 and x86_64 - the change makes it easier for developers to switch to a build with less testing and a smaller footprint.

Leverage Android skills and Gemma 4 in Android Studio Quail 4 (7 minute read)

Android Studio Quail comes preloaded with 23 curated skills, and users can create custom skills to extend Agent Mode with specialized experience and custom workflows.

Small Is Beautiful (18 minute read)

Small language models are catching up fast on reasoning tasks, but it will be a while before large language models are obsolete for these tasks.

Introducing agentic video understanding with Gemini (7 minute read)

Google's new agentic feature for video analysis cuts token consumption by up to 88%, reduces costs by up to 66%, and boosts quality by up to 7%.

Love TLDR? Tell your friends and get rewards!

Share your referral link below with friends to get free TLDR swag!

https://refer.tldr.tech/c463e0ba/

Track your referrals here.

Want to advertise in TLDR? 📰

If your company is interested in reaching an audience of tech executives, decision-makers and engineers, you may want to advertise with us.

Want to work at TLDR? 💼

Apply here, create your own role or send a friend's resume to and get $1k if we hire them! TLDR is one of Inc.'s Best Bootstrapped businesses of 2025.

If you have any comments or feedback, just respond to this email!

Thanks for reading,
Dan Ni & Stephen Flanders

Manage your subscriptions to our other newsletters on tech, startups, and programming. Or if TLDR isn't for you, please unsubscribe.

Fable 5.1 🤖, Waymo vs Tesla 🚕, agentic testing 👨‍💻