I build small, testable products at the boundary of computation and AI. Slid Phi Labs focuses on tools that make software more reliable, inspectable, and useful.
| Product | What it does | Link |
|---|---|---|
| CuNi | Write your code once, and it either behaves exactly the same in every language it builds for — or refuses to build at all. 144 languages and counting. | site · source · crates.io |
| Agent Rider | The live network where agents hold signed identities and pay per call. No signup, no API keys; the wallet is the auth. | site · source |
| Chamber JSON | A lockbox for secrets: seal API keys, credentials, and config into one encrypted package. Opening it takes two separate keys. | site |
| The Ring | Memory made of light — data as sustained laser interference patterns. Volatile by nature: it's RAM, not a disk. | live service · source |
| Tollkeeper | Seven small tolls at the chokepoints agents can't avoid: identity, discovery, escrow, grants, verification, bonds, memory. Caller pays; every toll returns a signed receipt. | site |
| AwLPay | Any chain, any token with verifiable value. Anything-to-anything conversion when a valid path survives fees — otherwise it refuses. | live · site |
| ExactOdds | Provably-fair gaming: commit-reveal outcomes with byte-identical rules across every runtime. The refusal is the product. | site · source |
| TRUSTREAM | Lossless compression for live data streams, in 4 KiB tiles. | site |
| Pulsar | An open GPLv3 lossless compressor and reproducible technical demonstration from Slid Phi Labs. | source |
| Quikgater | Pay-per-fact web fetching for AI agents using the x402 payment protocol. | source |
| Great Agentic Olympiad | An open competition framework for measuring agent and human–agent team performance. | project · site |
Reproducibility, honest benchmarks, and products that can be inspected. Public repositories contain selected demonstrations and open components; private research and hosted commercial systems remain separate.