Archived landing draft. Historical claims on this page require evidence review. Use atomik.tech, the current claims registry, and evidence labels for public claims.

Stop moving data.
Start evolving it.

ATOMiK evaluates whether state-aware execution can reduce wasted state movement in workloads constrained by bandwidth, latency, power, heat, or footprint. Claims require workload, artifact, context, and caveat.

$ pip install atomik-core
REFERENCE 0xDEADBEEF ACCUMULATOR 0x000000FF = CURRENT STATE 0xDEADBE10
Target
Bandwidth Evaluation
Proof
Formal Work
Case
Memory Evaluation
Measured
Artifact Required

How It Works

Four operations. Everything else is built on top.

LOAD

Set reference state,
clear accumulator
O(1)

ACCUM

XOR delta into
accumulator
O(1)

READ

Reconstruct state:
ref ⊕ acc
O(1)

SWAP

Atomic snapshot +
new epoch
O(1)
STATE RECONSTRUCTION — NOT VON NEUMANN TRADITIONAL: STORE FULL STATE Snapshot 1 64 KB Snapshot 2 64 KB Snapshot 3 64 KB ... 10,000 64 KB 655 MB memory for 10K rollback points ATOMiK: RECONSTRUCT FROM DELTAS Reference 8 bytes Accumulator 8 bytes 24 bytes for ANY number of rollback points

Why ATOMiK

Four operations. Evidence-labeled proof path.

BANDWIDTH

Traffic Reduction Target

Evaluate whether meaningful deltas can reduce full-state movement for a specific workload, baseline, and decision metric.

CONVERGENCE

No Consensus Protocol

XOR commutativity means deltas can arrive in any order. No Raft, no Paxos, no leader election. All nodes converge.

MEMORY

O(1) State Tracking

24 bytes per context regardless of history. 1 million checkpoints costs 24 bytes, not 8MB of snapshots.

CORRECTNESS

Formal Proof Work

Formal proof work exists for the algebraic model. Quote proof counts only from the current proof packet or claims registry.

DETECTION

Measured Change Detection

Change-detection results are workload-specific and must be quoted with the measured artifact, baseline, and caveat.

UPGRADE PATH

Software → Hardware

Same operation vocabulary across software and hardware paths. Quote speedups only from current proof artifacts.

Traditional vs ATOMiK

Side-by-side comparison for distributed state sync across 3 nodes.

Conventional Approach
Architecture Event sourcing + Raft
Per update Full state copy (64 KB)
10K updates bandwidth 655 MB
Rollback memory 8 MB (snapshots)
Change detection O(n) full rescan
Message ordering Required (consensus)
Correctness Tests + hope
vs
ATOMiK
Architecture XOR delta algebra
Per update 8-byte delta
10K updates bandwidth 80 KB
Rollback memory 24 bytes (always)
Change detection O(1) incremental
Message ordering Any order (XOR commutes)
Correctness Proof work present

4 operations.
That's it.

LOAD sets the reference. ACCUM XORs a delta. READ reconstructs the state. SWAP starts a new epoch. Everything else -- rollback, merge, fingerprinting -- is built on these four.

Zero dependencies
Python 3.9+
Fully typed
C99 header also available
from atomik_core import AtomikContext ctx = AtomikContext() ctx.load(0xDEADBEEF) ctx.accum(0x000000FF) # XOR delta state = ctx.read() # 0xDEADBE10 # Undo? Just re-apply the same delta. ctx.rollback(0x000000FF) assert ctx.read() == 0xDEADBEEF # Merge two independent accumulators. # Order doesn't matter (commutativity). a = AtomikContext() b = AtomikContext() a.load(0xCAFE); b.load(0xCAFE) a.accum(0x11); b.accum(0x22) a.merge(b) # a now has both deltas

Benchmarks

Review current benchmark artifacts before quoting speedups. Implementation tier, workload shape, and evidence label matter.

Network Bandwidth

64 KB state, 10,000 updates, 3 nodes
Full-state replication 655 MB
ATOMiK deltas 80 KB
8,192x reduction Same state, fraction of the bytes

Rollback Memory

1 million checkpoints for rollback
Snapshot stack 8 MB
ATOMiK (ref + acc) 24 bytes
case-specific memory comparison Constant regardless of history

Change Detection

1 MB buffer, incremental single-field update
SHA-256 full rehash 7,864 ms
ATOMiK fingerprint 6 ms
artifact-specific result O(1) incremental vs O(n) rescan

Multi-Node Convergence

8 nodes, 1,000 updates each
Event sourcing (sort + replay) Sort 8K events
ATOMiK (any order) XOR merge
order-independent algebra XOR commutativity = order-free

Same Algebra, Any Speed

Same API at every level. Upgrade when you need the throughput.

5M
Python
pip install atomik-core
500M
C Library
Single-header, any compiler
Artifact
FPGA
Quote only from current proof packet

Illustrative throughput tiers. Proof artifacts and implementation-specific validation are required before quoting performance or correctness claims.

Performance & Security

Deterministic model. Security claims require a reviewed threat model.

Software Performance

Metric ATOMiK Traditional Advantage
Throughput (Python) 5M ops/sec Varies by implementation pip install
Throughput (C) 500M ops/sec Varies by implementation Single header
Memory per rollback 24 bytes (constant) 8B × history depth Case-specific
Bandwidth per update 8 bytes (delta) Full state size 8,192× less
Time complexity Workload-dependent O(n) replay/scan Constant
Dependencies Zero Framework-dependent No bloat

🛡 Data Integrity & Security

Property Traditional ATOMiK
State tampering Undetected Requires measurement and threat model
Replay attacks Must track sequence numbers Algebraic cancellation; security use needs review
Data corruption Silent unless checksummed Can be evaluated with an integrity model
Convergence Requires consensus protocol Order-independent algebra for applicable deltas
Determinism Implementation-dependent Formal proof work present
Operation timing Variable (data-dependent) Security claim requires threat-model review
O(n)
Traditional
replay log / full-state scan
O(1)
ATOMiK
reconstruct from accumulator
Constant
regardless of history depth or state size

Evaluation Access

Start with a scoped evaluation, not a generic subscription plan.

Public Repository

Free
  • Full atomik-core library
  • Python + C implementations
  • 4-operation API
  • AtomikTable, DeltaStream, Fingerprint
  • Apache 2.0 (evaluation)
  • Community support
Get Started

Investor / Licensing Diligence

Request-basedstrategic review
  • Investor brief and proof-bound narrative
  • Hardware evidence map
  • Current validation gates
  • ASIC feasibility path
  • IP and licensing discussion
  • Partner diligence support
Request Diligence

For Investors

ATOMiK is a state-aware compute architecture moving from proof artifacts toward evaluated customer, design-partner, and licensing opportunities.

Proof
Formal Work
Artifacts
SW + HW Evidence
Labeled
Hardware Evidence
Scoped
Performance Claims
ICP
Edge / Embedded First

Business model: proof reviews, technical evaluations, design partnerships, and licensing/IP diligence. Pricing and margins require current business review.

Contact Founder Technical Deep Dive

Our Mission

We believe many constrained systems pay for unnecessary state movement. ATOMiK exists to evaluate where repeated scans, syncs, replay, and reconstruction create waste, then measure whether a state-aware architecture can reduce that waste for one workload, one baseline, and one constraint.

View on GitHub