Urgent.News

What's breaking now, across hundreds of outlets.

Tech

Interaction Nets Explained: The Programming Model Behind the Frostbite CPU

On November 18, 2026, the chip-design startup Architect Labs published a post introducing Frostbite . They describe it as a general-purpose CPU with a new instruction set. They say their AI systems designed it, from the compiler down to the final chip layout. The Hacker News thread reached 657 points and 342 comments. Most coverage is about the AI and the headline number: 4.59x better performance…

On November 18, 2026, Architect Labs, a chip-design startup, unveiled Frostbite, a new general-purpose CPU featuring an innovative instruction set. The AI-driven design process, from compiler to chip layout, prompted widespread interest in the technology. While most coverage focused on the AI aspect and Frostbite's headline performance figure – 4.59x better performance per watt than conventional cores – this article delves into the programming model behind Frostbite: interaction nets.

Interaction nets, a model of computation dating back over 35 years, enable Frostbite to optimize energy usage by preserving dependency information during the compilation process. This contrasts with traditional CPUs, which discard such information and linearize the program into a sequence of instructions.

Interaction nets operate on four core principles. First, programs are represented as graphs with nodes that have one principal port and zero or more auxiliary ports. Wires connect these ports. Second, computation occurs through rewriting pairs when a wire connects the principal ports of two nodes, replacing them with a new piece of graph per a defined rule.

Third, active pairs never overlap, as each node has only one principal port, ensuring independent rewritings and eliminating race conditions. Finally, every value is used exactly once, embodying the principle of linearity.

The Frostbite design achieves remarkable energy efficiency by leveraging interaction nets. Compared to a conventional CPU, Frostbite's quad (a two-by-two group of tiles) achieved a geometric mean of 4.59x better performance per watt, with 6.02x on tuning workloads and 3.83x on held-out workloads. This is primarily due to interaction nets' ability to keep dependency information, reducing the energy spent on memory, fetching, decoding, and instruction issuing – the most energy-intensive components in conventional CPUs.

Linearity, a key concept in interaction nets, is also integrated into the programming language Vine. Written in a syntax similar to Rust and compiled to interaction nets, Vine allows developers to write programs as interaction nets, making the energy-saving benefits more accessible. With its type system emphasizing "flexibility" through annotations like forkable and droppable, developers can easily manage resource usage and reduce unnecessary energy consumption.

Rewritten by urgent.news; may contain errors — check the original.

Read the original at dev.to →

Report a problem with this story

More in Tech

My --config-test accepted a listen address that could never bind

transponder is a Go proxy I've been building. It has a --config-test flag: load the config, validate it, exit 0 or 1. It runs in CI and in the deploy script before anything starts.

  • Configuration test passed despite invalid listen address ":nope"
  • Server crashed 200ms after passing --config-test
  • Author added rule to reject unspecified addresses with same port

The Fix That Brought the Bug Back

I use a second iPhone at home through a stream ( the setup ): the phone's screen comes off USB as H.264, an old laptop's GPU re-encodes it, and a fork of Sunshine streams it to my main phone.

  • Bug resurfaces after initial fix, causing connection delay and termination
  • Transcoder fixes include timeout for bus threads and replays for new viewers
  • System functionality restored with updated transcoder and fallback mechanism

You Don't Need That useEffect: Derived State in React

A search box over a list. It works, and it shows up in code review every week: function Results ( { items , query } ) { const [ filtered , setFiltered ] = useState ( items ) ; useEffect ( ( ) => {…

  • Using useEffect to derive state causes inefficiency in React components.
  • Filtered results stored in state create "effect-after-paint gap" on slow devices.
  • Compute values directly in component rendering to improve performance.

More from Saturday 21 November →