HigherOrderCO
Higher Order Company develops open-source programming-language and runtime technology intended to make high-level code fast on CPUs and GPUs while making AI-generated software more precise and verifiable. Its current Bend 2 positioning combines C-like speed, CUDA-style parallelism, Lean-like proofs, and Python-like syntax.
At a Glance
- AI-native software developers and teams
- Developers building high-performance backend systems
- GPU and parallel-computing developers
- Functional-programming and formal-methods researchers
- +1 more
AI Tools by HigherOrderCO
(1)
Bend
AI Language for Parallel Proofs
Discussions
No discussions yet
Be the first to start a discussion about HigherOrderCO
Latest News
Bend 2 launched as a proof-checking, GPU-capable programming language designed to block AI mistakes.
Bender launched as an AI agent specialized in Bend proofs; the initial version was described as a thin wrapper over public models, with SupGen integration planned.
Independent technical review documented Bend 2's launch, its LAWS/PROOF model, native CPU/GPU compilation, and planned paid Bender proving agent.
Bend 2.0.32 changelog/release shipped compiler, runtime, proof-verdict, platform, documentation, and test updates.
Products & Services
A fast, Python-syntax programming language with strong/dependent types, affine resource use, native C/Metal/CUDA/JavaScript compilation, automatic divide-and-conquer parallelism, and laws/proofs that can block code changes that violate declared invariants.
The original high-level, massively parallel language, designed to run Python-like programs over HVM/HVM2 across CPU cores and GPUs; it was later superseded by Bend 2.
A parallel interaction-net runtime / higher-order virtual machine for evaluating symbolic and functional programs, including GPU-oriented HVM-CUDA work.
A minimal, efficient functional programming language and proof checker/assistant developed by the team.
Market Position
Higher Order Company positions Bend 2 at the intersection of AI coding, formal verification, and high-performance parallel computing: Python-like ergonomics with native CPU/GPU compilation and explicit proof gates. Its closest technical comparators include Lean and Rocq for theorem/proof checking, CUDA for GPU parallelism, and Rust for resource-aware safety; Bend's differentiator is combining these ideas in a language aimed specifically at AI-directed software construction. The project is young and explicitly warns that it is still evolving.
Leadership
Founders
Victor Taelin
Brazilian functional-programming researcher; previously worked at the Ethereum Foundation, led development at the Kindelia Foundation, and created Formality and the Higher-Order Virtual Machine (HVM).
Executive Team
Victor Taelin
Founder and Tech Lead
Brazilian functional-programming researcher; previously associated with the Ethereum Foundation and Kindelia Foundation, and creator of Formality and HVM.
Nicolas Abril
Developer
Developer at Higher Order Company and contributor to the Bend repository; repository metadata identifies him as a contributor to current compiler and test work.
Founding Story
Victor Taelin says he started Higher Order Company in 2023 to turn HVM's massively parallel interaction-net research from a prototype into stable production technology and to bring high-level languages such as Python and Haskell toward lower-level performance. The initial plan was to ship HVM, use it as a compile target, and support symbolic and parallel applications.
Business Model
Revenue Model
The core language is intended to be 99% open source and usable offline for free. Generative features such as gen are planned to be metered through a paid API; Bender is offered as a paid proprietary proving agent, with credits usable across the ecosystem.
Pricing Tiers
Launch announcement said Bender credits were offered with a 50% founder discount; no standard price was stated.
Target Markets
- AI-native software developers and teams
- Developers building high-performance backend systems
- GPU and parallel-computing developers
- Functional-programming and formal-methods researchers
- Teams seeking mechanically checked invariants for AI-generated code
- AI-assisted and vibe-coded backend applications where invariants must not be violated
- High-performance CPU/GPU applications and massively parallel algorithms
- Symbolic AI and functional-programming runtimes
- Formalized application rules such as financial-balance invariants, bounds checks, game rules, and sorting properties
- GPU execution without hand-written CUDA kernels
- Purely functional game engines and modular execution layers