Tools
2026-08-12
Three years after promising Python's comfort with Rust's safety, Mojo calls itself 1.0 — with the compiler still closed
Modular released Mojo 1.0 on 11 August, the first stable release of a language it has been building since 2023. Mojo is the work of Chris Lattner, who created LLVM and Swift, and its purpose is narrower than it looks: one language that can target CPUs, GPUs, custom accelerators and whatever else an AI workload lands on, so that the stack does not have to be rewritten per chip. It reads like Python and borrows memory safety and systems-level control from Rust. The 1.0 label is mainly a promise about stability — changes through the 1.x line will be primarily additive, with breaking changes handled the way mature languages handle them. The release itself adds Python-style lambda syntax, a steadier language server for editors, memory-safety diagnostics for invalidated references, and a unification of closures and pointer types. The standard library has been open source since 2024, with nearly 200 contributors landing more than 1,100 pull requests across 200,000 lines; the compiler and toolchain are not open, and Modular repeats its pledge to open them during 2026.
Why it mattersThe reason a language like this exists is that AI code today is written twice: once in Python, for people, and once in CUDA or something like it, for the machine. That split is why so much AI software is locked to one vendor's hardware — the fast part is written in a dialect that only runs on one company's chips. Mojo is a bet that the two halves can be one language, which if it works is a slow, structural loosening of Nvidia's grip. The caveat is the compiler. A language whose toolchain is controlled by a single company is asking developers to trust a roadmap, and 1.0 arrives with that promise still unkept.
✓ Verified · 3 sources
Read in the app — free, in 9 languages
Related stories
Apple Music will tell you when a song was made by a machine — but the uploader decides whether to say so
2026-08-21Rumour: the anonymous model that just topped a coding benchmark, for free, is said to be Zhipu's unreleased flagship
2026-08-21Nvidia is paying $6 billion for the machine that builds a rival's models — and hiring 109 of the people who ran it
2026-08-21Stripe has just paid $7.5 billion for a model router. Days later Ramp built one and is giving it away until January.
2026-08-21Meta's assistant is now a Mac app that reads your screen and types into any window — and what it sees can train the model
2026-08-21