The Verifiable Agent
Why autonomous agents need verifiable execution, trustworthy memory, and a provable right to act.
Hi, I’m Nico. A technical founder building AI products for businesses, agent systems, and the infrastructure behind them.

Helping companies license their workplace data to AI labs. Data is de-identified, and the company decides who can buy it.
Visit grokked.aiAI for recovering overdue invoices through conversations, payment agreements, and follow-up. Collection fees are payable when money is recovered.
Visit kollekt.nowI co-authored this work on checking that a remote language model ran as intended, without running the whole model again on the verifier’s machine.
I’m currently extending TensorCommitments to verifiable post-training using lattice-based commitments, with a focus on performance and post-quantum security.
Accepted at NeurIPS 2026
Co-authored with researchers from UT Austin, Georgia Tech, and McGill.
Start with one token at a time, then explore parallel guesses, the training that makes them useful, and what the speed and accuracy numbers actually show. Includes an animated explanation.
Read the articleWhy autonomous agents need verifiable execution, trustworthy memory, and a provable right to act.
How agents extend programmable execution with the ability to interpret context and make decisions.
A critique of the economic incentives between Ethereum and its rollups.
A practical account of the hardware, setup, and results of running a Livepeer node.
Building a simple CRUD API to explore Go, gRPC, and MongoDB.
A code-led explanation of proxy contracts and delegatecall, with a working example.
Head of Protocol & Infrastructure
August 2026 to present
Leading protocol architecture and infrastructure for people, AI agents, and robots. Building user-owned personal AI context, private payments, and a naming protocol for a Sui-derived chain.
Built a self-balancing peer-to-peer GPU network for LLM inference and PEFT/LoRA post-training, adapting Qwen3 and DeepSeek-R1 for distributed execution. The app won a prize at the 2024 DMCC × Bybit hackathon.
Owned architecture and most of the code, and led three engineers and the executive team. Built validator-specific liquid staking across 150+ validator destinations and TenderSwap, which reached $10M in cumulative swap volume.
Founder & Engineer
February 2020 to December 2023
Built, operated, and open-sourced a GPU video-transcoding pool. The software was adopted by major Livepeer operators, including the official grants node.
An early engineering hire in a two-person protocol team. Co-designed probabilistic micropayments, owned the orchestrator/transcoder split, and guided the pre-launch Arbitrum migration through design and review. In 2024, built open-source Livepeer AI infrastructure for LLMs, video models, and object detection.
Blockchain Lead
May 2018 to May 2019
Established Emakina’s new blockchain business unit and led architecture, prototyping, and enterprise and government pre-sales. Delivered pilots in telco settlement, trade finance, identity, and supply-chain provenance. Won the Emakina Group Silver Innovation Award.
My work spans applied AI products, agent systems, distributed compute, and protocols. I like working across the whole stack: the model, the runtime, the infrastructure, and the product people actually use.
My work ranges from early protocol engineering at Livepeer and founding Tenderize to agent compilers, memory systems, private personal AI, and distributed inference and post-training.
I’m based in Belgium and have worked remotely with teams in New York, San Francisco, and worldwide since 2019. I speak Dutch and English, with conversational French.
If you’re building applied AI products, agent systems, or distributed compute, I’d like to hear what you’re working on.
Message me on LinkedIn