The Pulse #127: Typescript compiler rewritten in Go
Why did Microsoft choose a language developed by Google, instead of its own, C#? Also: OpenAI’s dev tools for agents looking complex, the PR/FAQ for AWS Lambda, and more
The Pulse is a series covering insights, patterns, and trends within Big Tech and startups. Notice an interesting event or trend? Send me a message.
Today, we cover:
Industry pulse. Google’s impressive AI launches, AI agent Manus makes a splash, companies evaluating AI models internally, Cursor worth $10B, Klarna’s upcoming IPO, and more.
Typescript compiler rewritten in Go. In an impressive feat, the Typescript team rewrote their compiler, and made it 10x faster. But why use Go for the job? It’s a curious choice, given Microsoft’s heavy investment in C# – which is also a performant language. Engineers who executed the rewrite explain their unconventional choice.
OpenAI’s dev tools to build agents: more complexity? OpenAI shared its tooling for building AI agents. My first impressions are that it feels pretty fragmented. Perhaps this is to be expected with fast-moving technology, where capabilities keep being added?
Apple and AI: clumsy fumbling, or a long-term strategy? Apple promised a new, improved Siri by the end of 2024, yet that could now be on hold until 2027. Are things less bad as they seem, due to Apple owning the iPhone hardware and iOS software platforms?
Second chances for interview no-shows? A candidate did not show up for a product engineer interview, but their AI note taker did. As a hiring manager, what would you do? Cofounder Diwakar Kaushik did something unexpected.
Industry Pulse
Google’s impressive AI launches
Google has had an eventful weeks with new AI products:
Gemma 3: A model that runs on just a single NVIDIA H100 GPU, and requires around 10x less compute than models with similar capability, is impressive – and will surely gain adoption to its generally permissive license, and can be self hosted, as well as supporting 140 languages (!!). As a note, Gemma 3 is kind of open, but not fully open: the license allows commercial use, but has restrictions on use cases, such as violating any applicable law or use cases listed in the policy.
Gemini Robotics: a vision-language-action model that can directly control robots, built on top of Gemini 2.0. This 3-minute video summarizes the impressive capabilities.
Google is getting a lot less of the AI spotlight behind OpenAI and Anthropic, but seems to be pulling ahead in other areas. In robotics and AI, it might become pioneering, and there’s also Waymo's self-driving cars – which is also part of Google. Is Google quietly becoming a ML and AI leader under the radar?