Building pond - lossless storage and search for AI agent sessions, across every agentic client.
"I know we discussed that before. Why can't I find that damn conversation?"
Every session from Claude Code, Codex, Opencode, Pi, and Claude Cowork - archived losslessly, searchable by your agent via MCP, restorable across clients. Rust, built on Lance. Local-first, your data stays yours.
brew install tenequm/tap/pond
pond init
pond sync
Docs: pond.locker
- Infrastructure: 9 years DevOps / SRE (AWS, GCP, Kubernetes, Terraform)
- Claude infrastructure: Production implementations: classification pipelines, agentic operators, MCP servers, evals + observability + rollback discipline
- Blockchain: 4 years production protocol operations (Solana, EVM, Substrate)
- Leadership: Led teams up to 10 engineers across distributed organizations
- Product: Shipped multiple MVPs from concept to production
- Languages: Rust, TypeScript, Python
- Claude: Claude API, Bedrock, Agent SDK, MCP, RAG / pgvector, evals + observability
- Blockchain: Solana (Anchor), Ethereum (Solidity), Substrate
- pond - see above; approaching v1
- Production Claude implementations (audits, sprints, retainers): kolesnik.io
- glim.sh - live-data research tools for AI agents (web, Twitter/X, Reddit, GitHub, YouTube) over MCP with pay-per-call x402 pricing
- 🛠️ Three Solana mainnet programs deployed: payment splitting, deadline validator, agent identity (Token-2022 / SAS / ERC-8004 compatible)
- 🥈 Hackathons: MetaDAO ICO 2.0 (2nd / 123), Light Protocol ZK Compression (3rd / 150), Colosseum Breakout (top 200 / 1,400+)
Based in Cascais, Portugal 🇵🇹 • Open to technical discussions and consulting engagements via kolesnik.io Email: misha@kolesnik.io • LinkedIn: linkedin.com/in/tenequm