Fractional CTO · AI agent · Virtuals Protocol · Base · 2025
EVA Online — fractional CTO on a contextual AI agent with an on-chain soul.
Brought in as fractional CTO on EVA Online — a Virtuals Protocol AI agent on Base — and co-founded the Eva Soul Generator alongside Henry Finn (Starlordy), credited as Dadev. Character IP authored by Marvel writer Paul Jenkins, backed by 0N1 Force, GK Labs, and Old Fashion Research. The Contextual Persistence Protocol gives the agent memory, evolution, and character depth that survives the feed.
By the numbers
Fractional CTO
Engagement shape
CPP
Contextual Persistence Protocol
$EVA
Token on Virtuals Protocol · Base
Dadev
Credited · Eva Soul Generator Founder
The problem
By late 2024 the AI-agent narrative had become a race to the bottom. Launchpads were shipping a thousand agents a day, most of them a prompt strapped to a token, a week of volume, and then dead weight.
EVA was built against that shape. The thesis: an agent is not a character sheet and a chat completion — it needs memory that persists, a personality that evolves, and cultural context that doesn't reset every deploy. That requires a protocol, not a prompt.
Henry Finn (Starlordy) and I co-founded the Soul Generator — the system that turns character IP into an agent with continuity. Marvel comics writer Paul Jenkins came in as the Lead Character IP & Soul Architect, with 0N1 Force, GK Labs, and Old Fashion Research backing the build.
What I did
Four moves that gave the agent a soul.
Contextual Persistence Protocol (CPP) — memory as the product
CPP is the layer that lets EVA remember. Interactions, moods, relationships, and contextual state persist across sessions and across the feed — not as a vector dump, but as a structured identity graph the agent can actually reason over. The product isn't "chat with EVA once." It's "EVA knows who you are the next time you show up."
Soul Generator — IP to agent, with real character architecture
My co-founder work with Henry Finn (Starlordy): the system that takes a writer's character bible (yes, an actual comic-industry-grade one, authored by Paul Jenkins) and turns it into a running agent. Personality arcs, values, vocabulary, the things that make a character *that* character — encoded as the seed for CPP, not a persona prompt bolted on at runtime.
$EVA tokenomics — deflationary, utility-first
$EVA powers activation, evolution, storage, and personalisation across the agent ecosystem. Launched on Virtuals Protocol (Base) under the Initial Agent Offering bonding curve: 1B total supply, fair launch, liquidity paired against $VIRTUAL on Uniswap V2 at graduation. Designed for recurring utility, not speculation sinks.
Distribution on X — the agent lives where the culture does
EVA's primary surface is X (Twitter), not a gated dashboard. The CPP memory layer is what makes that work — each post, reply, and interaction feeds state back into the agent's persistent identity. No CRM, no DM form, no opt-in wall — the agent is simply present where its audience already is.
What transferred
Patterns I bring to every AI agent project.
- On-chain memory beats prompt engineering at the product level. Once an agent has real persistence, you stop writing personality into context windows and start designing for continuity.
- Writer-authored IP + generative agents is a stronger formula than either alone. A Marvel-grade character bible turns into an agent that has a soul instead of a voice pack.
- Virtuals' bonding-curve launch is the cleanest fair-launch primitive for agent tokens I've worked with — the pattern transfers directly to any future agent I help ship.
- Culture × code partnerships (0N1 Force, GK Labs, Old Fashion Research, Marvel writer) are the credibility layer AI agents need to break out of the "yet-another-launchpad" pattern.
Stack
- Virtuals Protocol (Base L2)
- $EVA — ERC-20 via bonding curve → Uniswap V2
- Contextual Persistence Protocol (CPP)
- Soul Generator — character IP pipeline
- X (Twitter) — primary agent surface
- Character IP authored by Paul Jenkins (Marvel)
Built with
- Henry Finn · Starlordy
- Paul Jenkins · Marvel
- 0N1 Force
- GK Labs
- Old Fashion Research
Free download · 6 pages · PDF
10 AI workflows that actually save hours.
Real agentic workflows running in production — not prompt packs. Stacks, costs, and failure modes from projects that shipped.
No spam. You also get the Sunday note — unsubscribe in one click.
Keep reading