Bend
A new programming language designed for the AI era, featuring formal proof mechanisms for intercepting logical errors in AI-generated code and native GPU parallel execution via the HVM runtime. Its underlying computational model is based on interaction combinators.
Timeline (last 90 days)
Bend released a global interaction and collaboration layer product this week, supporting any agent, platform, language, and environment, with multi-point collaboration, a built-in registry for automatic discovery, persistence, channels, message filtering, security and observability, and support for human in the loop.
Vlad is the co-founder and CTO of Bend.
Bend introduces a file called laws.bend where rules are declared once and proven, and if the code is modified in a way that breaks that proof, the compiler rejects it before it goes live.
Bend runs at speeds close to C on CPU and with CUDA on GPU, proof checking completes within one second, and it can distribute work across thousands of cores without requiring hand-written threads or locks.
The Bend language was developed by the team at HigherOrderCO.
Bend is a programming language that aims to enable ordinary developers to write massively parallel programs.
Bend's core technology is based on the Interaction Combinators computational model, which can automatically map functional-style code to GPUs for massively parallel execution.
Bend allows developers to avoid explicitly managing threads, locks, or memory layout, as the compiler automatically infers parallel structures.
Bend employs a formal proof mechanism where the compiler verifies whether the proof holds, and if the AI-generated implementation does not conform to the specification, compilation fails, thereby intercepting errors at the compilation stage.
Bend's GPU execution capability is reportedly based on its underlying runtime HVM (Higher-order Virtual Machine).
3 more timeline events
All Facts (13)
Vlad is the co-founder and CTO of Bend.
50%UnverifiedBend introduces a file called laws.bend where rules are declared once and proven, and if the code is modified in a way that breaks that proof, the compiler rejects it before it goes live.
50%UnverifiedBend runs at speeds close to C on CPU and with CUDA on GPU, proof checking completes within one second, and it can distribute work across thousands of cores without requiring hand-written threads or locks.
50%UnverifiedBend's core technology is based on the Interaction Combinators computational model, which can automatically map functional-style code to GPUs for massively parallel execution.
50%UnverifiedBend allows developers to avoid explicitly managing threads, locks, or memory layout, as the compiler automatically infers parallel structures.
50%UnverifiedBend is a programming language that aims to enable ordinary developers to write massively parallel programs.
50%UnverifiedThe Bend language was developed by the team at HigherOrderCO.
50%UnverifiedBend is a programming language whose core selling point is intercepting errors in AI-generated code through formal proofs, while natively running on GPUs for large-scale parallel performance.
50%UnverifiedBend employs a formal proof mechanism where the compiler verifies whether the proof holds, and if the AI-generated implementation does not conform to the specification, compilation fails, thereby intercepting errors at the compilation stage.
50%UnverifiedBend's approach to proofs is in the same vein as dependent type languages (such as Idris, Lean, and Coq), but targets the specific scenario of guaranteeing the correctness of AI-generated code.
50%UnverifiedBend's design goal is to let developers write ordinary high-level code, with the runtime automatically discovering parallelism and mapping it to GPU execution, without the need to manually rewrite code in CUDA.
50%UnverifiedBend's GPU execution capability is reportedly based on its underlying runtime HVM (Higher-order Virtual Machine).
50%UnverifiedBend released a global interaction and collaboration layer product this week, supporting any agent, platform, language, and environment, with multi-point collaboration, a built-in registry for automatic discovery, persistence, channels, message filtering, security and observability, and support for human in the loop.
50%