TLDR 2026-03-17

Attending NVIDIA GTC, virtually or in-person? See what Google x Nvidia have been cooking at these hands-on sessions:

• Discover how Google Cloud's AI Hypercomputer with NVIDIA Blackwell GPUs helped leaders like WPP, General Motors and Schrödinger build their AI stacks.
• Hear from the product leads on the technical capabilities of the Google Cloud NVIDIA Blackwell systems
• Explore how agents and digital humans are redefining customer interactions at Puma

P.S. Don't forget to stop by the Google Cloud booth for some goodies and to hear the latest updates directly from the Google team!

Nvidia introduced a huge array of new products at its annual GTC conference on Monday, ushering in new hardware and software geared towards running AI models more quickly and efficiently. The company has been signaling for most of the last year that it would increasingly focus on inference computing going forward. Its traditional GPUs are not ideal for inference because they consume a large amount of energy and don't come with enough attached memory. Nvidia expects to sell $1 trillion worth of its Blackwell and Rubin chips, focused on inference, by the end of 2027.

OpenAI's 'do everything at once' strategy has put the company on the defensive. Its top executives are finalizing plans for a major strategy shift that will refocus OpenAI around coding and business users. The company's leaders are actively looking for areas to deprioritize. OpenAI has taken steps towards an initial public offering that could open in the fourth quarter of this year.

The entire neuronal diagram of the fruit-fly brain was mapped in 2024. Later research showed the map can be used to predict behavior. The brain map has now been uploaded into a digital environment. It's not a simulation - the fly's 'sensors' are being activated by the digital environment and the neurons are responding. The digital fly appears to behave in the digital environment in reasonably fly-like ways. A video that explains the yet-published experiment (with footage of the digital fly) is available in the article.

Nvidia's Vera Rubin Space-1 Module is designed for orbital data centers. Specifically engineered for size-, weight-, and power-constrained environments, the chips will be used on space missions led by multiple companies. There are still engineering hurdles to overcome in building orbital data centers. For example, in space, there's no convection, so Nvidia has to figure out how to cool the system with just radiation.

AI has broken the language barrier. What comes next? See how Google Cloud and NVIDIA are transforming robotics in this exclusive session with Disney, NVIDIA, DeepMind and Skild AI; or tune in to a fireside chat between DeepMind Chief Scientist Jeff Dean and NVIDIA Chief Scientist Bill Dalyon the next frontier of AI.

Better coding tools produce better models, which produce better coding tools. The bottleneck has moved to managing long-horizon tasks and deploying agents in regulated settings. The human role is shifting to planning, evaluating, and steering. Enterprise adoption is gated by permissioning, sandboxing, and regulatory caution more than by model capability. Slop and convergence are real concerns, and context remains unsolved.

Investments in developer productivity can provide unexpected dividends when agents are included in the workflow. Stripe's Minions work in isolated environments with hybrid orchestration that mixes deterministic guardrails with agentic flexibility. Curated context needs agents with the right information without overwhelming them, and fast feedback loops help to iterate. Each layer is necessary, and none alone is sufficient.

Apple's AirPods Max 2 features an updated H2 chip that makes active noise canceling up to 1.5 times more effective compared to the original AirPods Max. It should also make transparency mode more natural due to a new digital signal processing algorithm. The H2 chip also adds support for adaptive audio and Conversation Awareness, a feature that automatically lowers the volume when someone starts talking to you. It also adds support for live translation and voice isolation. The AirPods Max 2 will be available for preorder on March 25 for an early April launch for $549.

Nvidia has unveiled a new software toolkit that helps businesses more safely build and run autonomous agents with the OpenClaw platform. OpenClaw allows developers to build autonomous agents that can plan and execute tasks on their own using subagents that can tackle specialized tasks, access local file systems, and delegate. Nvidia's NemoClaw is designed to help agents run safely in an enterprise context within contained virtual environments. Nvidia is working with OpenClaw's creator to make agents more trustworthy, scalable, and accessible.

Companies that aren't seeing increases in revenue despite productivity improvements could be oversized and may need to adjust revenue multiples accordingly.

Atlassian is restructuring following a plunge in its stock price driven by developments in artificial intelligence.

GLP-1s are made up of a chain of amino acids chemically bonded together, which just looks like food to the body, making it hard to create an oral formulation.

Biosecurity is one of the few areas well-suited to the game of venture capital, but it needs a catalyzing event.

Fueled by a 1.5X increase in IPC from its Olympus cores and an innovative high-bandwidth design, the new Vera CPUs claim impressive 50% performance gains over standard CPUs.

As the first engineer on TLDR's new Applied AI team, you'll build AI agents to automate end-to-end workflows and lead our context engineering project to allow non-technical teammates create their own AI workflows. Learn more.

Google is rolling out Android developer verification in the coming weeks, requiring app makers outside of the Play Store to register with their real names and pay a fee to Google or have their apps blocked from installation on virtually all Android devices.