Frequently Asked Questions

ATOMiK is a delta-state algebra engine that fundamentally departs from Von Neumann architecture. Instead of storing and retrieving state, ATOMiK reconstructs it: current_state = initial_state XOR accumulator. The underlying math forms an Abelian group (commutative, associative, self-inverse, with identity), rigorously proven through 92 Lean4 theorems.

Traditional computing copies full state on every operation, moving enormous amounts of redundant data. ATOMiK eliminates this by tracking only the changes (deltas) between states. This reduces memory traffic by 7,670x to 916,000x across validated workloads, with 22–58% execution time improvements.

The core SDK is open source under the Apache 2.0 license. The kernel module and enterprise features (FPGA IP, hardware acceleration, priority support) are commercially licensed. See our pricing page for details.

ATOMiK provides SDKs for Python, C, JavaScript, Rust, and Verilog, all generated from a single canonical specification via the SDK generator pipeline. Install the Python SDK with pip install atomik-core.

Yes. The free tier is built on the Apache 2.0 licensed core SDK, which has no restrictions on production use. You can deploy ATOMiK in production applications at no cost.

When your Pro trial ends, trial-exclusive features gracefully disable while your data is fully preserved. You can upgrade to a paid Pro plan, downgrade to the free tier, or contact sales for Enterprise options — all without losing any work.

Yes. FPGA IP cores, hardware acceleration primitives, and the Verilog integration layer require an Enterprise license. This includes the ATOMiK parallel-bank core validated at up to 69.7 Gops/s on Xilinx Zynq.

We offer academic licensing for qualified institutions and research programs. Contact sales@atomik.tech with your institution details for pricing.

The SDK works everywhere Python, C, or JavaScript runs — Linux, macOS, and Windows. The kernel module requires Linux 5.15 or later. FPGA integration is supported on Gowin (Tang Nano 9K) and Xilinx (Zynq 7000 series).

Both ATOMiK and CRDTs enable conflict-free convergence in distributed systems. The key difference is algorithmic complexity: ATOMiK uses O(1) XOR algebra for merges, while CRDTs typically require O(n) merge functions. This makes ATOMiK significantly faster at scale. See our detailed comparison for benchmarks.

View comparison →

Near-zero. The Python SDK adds approximately 200 nanoseconds per operation. The C library runs at about 2 nanoseconds per operation. On FPGA hardware, the parallel-bank architecture achieves 69.7 Gops/s. Overhead for tracked memcpy is just 5–12% versus plain memcpy.

See benchmarks →

Security in ATOMiK is architectural, not bolted on. The design provides deterministic latency (eliminating timing side channels), has no speculative execution (eliminating Spectre/Meltdown class attacks), and no cache coherency (eliminating cache-based covert channels). Dynamic reference states provide additional protection.

View compliance details →

Support tiers match your plan: Community users get help via GitHub Discussions. Pro subscribers receive 48-hour email SLA support. Enterprise customers get 4-hour SLA with a dedicated support channel and named account engineer.

File bug reports on GitHub Issues. Include your SDK version, language, OS, and a minimal reproduction case for fastest resolution.

Open an issue

Start a thread on GitHub Discussions for community input, or email support@atomik.tech for direct requests. Enterprise customers can request features through their dedicated support channel.