# === / ===
# source: frontend/public/index.md

---
title: Jesse Anderson
description: 30-year coder. I build platforms. Currently obsessed with AI infrastructure.
url: /
---

# Jesse Anderson

> Bay Area, California.

**30-year coder. I build platforms. Currently obsessed with AI infrastructure.**

## Three audiences. Same site.

- **Recruiters & clients** — Real systems in production, real scale, real doctrine. See [/work](/work).
- **Cardano & AI peers** — Concrete projects, actual stack, the AI-infra specifics. [/now](/now) is what I'm deep in this month.
- **The curious** — Non-linear arc, immersion pattern, four boys, two granddaughters. [/about](/about).

## Current immersion

AI as real infrastructure. Not autocomplete. K.O.R.A. — autonomous Discord ops agent for a production Cardano platform. Multi-model fallback chains. Dual-GPU local LLMs. A Codex CLI fork. Set-me-loose-all-night runs as a working mode. More on [/now](/now).

---

Bay Area. Father of four. Coaches the youngest two on baseball. Built an AI bridge through OpenClaw so his third son can direct 3D scenes by voice. [This site is also addressable by agents](/standards).


# === /work ===
# source: frontend/src/pages/work.mdx

---
layout: ../layouts/Base.astro
title: Work — Jesse Anderson
description: ADA Handles, HAL, K.O.R.A., OpenClaw, Papagoose, AI tooling. The work shipped.
slug: work
---

<main class="work">

<PageHeader eyebrow="Work" title="The work shipped.">
Three peer products on Cardano — ADA Handles, HAL, and K.O.R.A. — plus side products and earlier engineering. Most of [github.com/koralabs](https://github.com/koralabs) is mine, outside of front-end work that the frontend team owns. The personal AI tooling that runs alongside this lives on [/projects](/projects).
</PageHeader>

<Project eyebrow="ADA Handles" title="Custom wallet addresses on Cardano, secured on-chain." featured={true}>

Co-founded with my second-eldest son in 2021. I run platform engineering and the AI ops layer.

<ProjectStats stats={[
    ['265K+', 'address-resolution NFTs'],
    ['70K+', 'wallets resolved'],
    ['10K', 'NFT collection with updateable mint/edit'],
    ['All', 'major Cardano wallets adopted']
]} />

The platform is a constellation of services across the [`handle.me`](https://handle.me) namespace. The work in impact terms:

- **Killed the centralized mint server.** Anyone can mint or edit a handle on-chain within the contract's rules. Every major Cardano wallet adopted the pattern. *(`decentralized-minting`, `handles-personalization`, `cip-68-444-minting`, `cip-68-222-minting`.)*
- **The withdraw-0 / observer trick — contracts that upgrade in place.** Worked out the Cardano technique that lets a live contract have its logic upgraded without moving the locked UTxOs to a new address. This is what makes our cutover work without state migration or chain downtime. *(Currently shipping live in `handles-personalization`.)*
- **Two parallel cutovers off Helios.** Smart contracts moving to **Aiken** for type safety and auditability — DeMi and SubHandles are on Aiken; Personalization not yet. The TX-building library moving to **`@cardano-sdk/*`**.
- **OAuth-over-NFT-ownership.** A wallet holding a handle can act as an OAuth identity provider — Web2 SSO ergonomics on self-custody wallets, no custodian. I pioneered Lit at Kora Labs in this codebase; the patterns echo through to my [`lambda-site-template`](https://github.com/papag00se/lambda-site-template) (which this site is built on). *(`auth.handle.me`.)*
- **Merkle Patricia Tries inside on-chain code.** Ported a non-trivial cryptographic structure into a Cardano-targeting language so contracts can prove membership over large datasets without storing them on-chain. *(`merkle-patricia-forestry-helios`.)*
- **Browserless dynamic SVG rendering at scale.** Every handle's NFT image is generated server-side via typographic SVG — hundreds of thousands of unique renders across reissues and personalization. *(`handle-svg`, `render.handle.me`.)*
- **Cardano Node on AWS Fargate (ran in production; retired for cost).** `cardano-node`, Ogmios, and the Handles API on Fargate + EFS for redundancy, failover, auto-scaling, and reproducible deployment. The fun trick: cross-host node-socket connectivity solved via `socat` bridging Unix domain sockets to TCP. Production-grade serverless ops on a chain that didn't expect to be run that way. Eventually retired — Fargate at this scale wasn't sustainable on the bill — but the technique still travels.
- **IPFS content validation inside on-chain code.** A Cardano smart contract can validate the contents of an IPFS file by matching the CIDv1 embedded hash against redeemer-supplied bytes — with care around chunk size and the ~1KB–12KB sweet spot. Live in production for personalization.
- **Discord as the production control plane.** Tickets, deploys, alerts, and ban decisions flow through Discord. Built the bot, the CloudWatch → Discord alerting pipeline, and the deploy automation operators trigger from chat threads. *(See K.O.R.A. below for the autonomous-agent layer on top.)*

<ImagePlaceholder note="ADA Handle visual — typography-led NFT grid" />

</Project>

<Project eyebrow="HAL — Handle Augmented Lifeform" title="One of the best minting experiences on Cardano." featured={true}>

A second product line built on the on-chain primitives Handles introduced. SDK + minting engine + 3D-asset support, so external Cardano teams can ship NFT collections without rebuilding the contracts and orchestration we already battle-tested.

<ProjectStats stats={[
    ['~4K', 'transactions chained in a single run'],
    ['70K+', 'tokens minted'],
    ['8 / 9', 'scripts and ref inputs per tx'],
    ['100%', 'deterministic settlement']
]} />

**The single most validating run (June 2025):** chained ~4,000 transactions minting 70K+ tokens in a single coordinated session. Every tx relied on the previous, with 8 scripts and 9 reference inputs each. *We were minting faster than the chain settled.* Every transaction landed deterministically.

For Cardano peers: this is what production-grade minting looks like when the contract chain, the engine, and the SDK are all yours. The minting infrastructure unified back to power Handles minting too — [`minting.handle.me`](https://minting.handle.me) serves both. *(Repo cluster: `hal.handle.me`, `hal-minting-engine`, `hal-minting-contracts`, `hal-sdk`, `hal-glb`, `minting.handle.me`.)*

<ImagePlaceholder note="HAL minting — chained-transaction visualization or 3D asset render" />

</Project>

<Project eyebrow="K.O.R.A. — Knowledge and Operations Response Assistant" title="Autonomous Discord ops agent for production Cardano.">

A multi-component AI agent that runs day-to-day ops for the Handles ecosystem from inside Discord — ticket triage, moderation, monitoring, alert response, deployment orchestration, knowledge synthesis.

**Architecture:** chat gate → AI provider → executor → worker. Multi-model routing with graceful fallback: cloud primary → codex-spark on usage cap → gpt-5.4-mini → local Qwen3 on dual GPUs (4B for the gate, 9B-IQ4_XS for the worker, 35B for harder reasoning when there's headroom).

**Discipline:** three-tier test battery — hermetic (mocks) → real-process (test services) → live-canary (live data, no side effects). Nothing ships green-by-inference.

**Working mode:** *set me loose all night.* TODO + TASK_STATE files keep autonomous runs resumable across crashes, compactions, and OAuth expirations. The agent rehydrates from disk.

**Discord-as-control-plane is intentional.** Pinned messages in `#kora-settings` are configuration. App-alerts post to `#app-alerts` with rich metadata for triage. Natural-language commands ("deploy to mainnet," "troubleshoot this ticket") trigger real CI/CD and GitHub work. *(Repo: `kora-bot`. Plus `cloudwatch-to-discord`, `community-jobs`. Detail on [/projects](/projects).)*

<ImagePlaceholder note="K.O.R.A. — Discord ticket flow, screenshot of operator action" />

</Project>

<Project eyebrow="Papagoose LLC" title="Side products.">

[clicked.bio](https://clicked.bio) — a LinkTree alternative with Instagram integration. [remindsly.com](https://remindsly.com) — Microsoft Graph-based reminders.

</Project>

<Project eyebrow="Other Kora Labs surfaces" title="Also shipping.">

A few more things in the [github.com/koralabs](https://github.com/koralabs) constellation worth knowing about:

- **Marketplace** ([marketplace.handle.me](https://marketplace.handle.me)) — secondary trading for Handles and HAL, on-chain order escrow.
- **HandleChat** ([chat.handle.me](https://chat.handle.me)) — direct messaging on Handle identity.
- **Social Secrets Restore** ([secrets.handle.me](https://secrets.handle.me)) — account recovery via social trust graphs. Research stage.
- **Merch** ([merch.handle.me](https://merch.handle.me)) and **Metaverse** ([metaverse.handle.me](https://metaverse.handle.me)) — announced.

</Project>

<PriorWork lede="Earlier engineering at scale, before Kora Labs.">

### Capital One — Master Platform Engineer (2018–2022)

25K records/sec synthetic-data Kafka platform on Fargate autoscaling. Constant vulnerability monitoring + threat scoring across team systems.

### Xero — Senior SWE & DevOps (2013–2018)

100+ two-week sprints on a 2M-user SaaS accounting/payroll platform. Built contextual auto-complete search over 2M objects on AWS ElasticSearch, queried dozens of times per second.

</PriorWork>

</main>


# === /projects ===
# source: frontend/src/pages/projects.mdx

---
layout: ../layouts/Base.astro
title: Projects — Jesse Anderson
description: Personal AI tooling, the rig, OpenClaw, the operating doctrine. Peer-leaning detail.
slug: projects
---

<main class="projects">

<PageHeader eyebrow="Projects" title="The rig, the tools, the doctrine.">
Higher detail than [/work](/work). The personal AI tooling layer — local LLM infra, the Codex CLI fork, multi-agent routing, the OpenClaw bridges, and the operating doctrine. Aimed at peers who care how the sausage is made.
</PageHeader>

<Project eyebrow="codex-local" title="What 9B can actually do." featured={true}>

The reason I'm bullish on local models: **codex-local** — my fork of the Codex CLI, tuned to babysit a local model through a structured task list — drives multi-turn, multi-file, multi-task coding sessions to completion using just **Qwen3.5-9B-IQ4_XS**.

<ProjectStats stats={[
    ['9B', 'parameters (Qwen3.5 IQ4_XS)'],
    ['8+', 'turns of sustained coherence'],
    ['4+', 'files edited per session'],
    ['Full', 'TODO list completion end-to-end']
]} />

The trick isn't the model — it's the scaffolding. Specifically:

- **Docs-first.** Every repo has an `AGENTS.md` and a `TASK_STATE.md`. The model reads them on every turn. Persistent intent survives compaction.
- **State in files, not in-context.** Long runs survive crashes, context exhaustion, model swaps, OAuth expirations. The agent rehydrates from disk.
- **Explicit task tracking.** TODO list in a file, model crosses items off as it goes. No "where was I" questions.
- **Verify before declare-done.** No green-by-inference. Run the test, read the output, look at the deployed version.

Most local-model coding agents stall past trivial single-edit tasks. With those four disciplines, 9B handles real work. The frontier model is no longer the moat — the scaffolding is.

The fork itself adjusted compaction behavior, model-routing interception, and exec-log inspection so a long unattended run could survive overnight without losing the thread.

</Project>

<Project eyebrow="The rig" title="Dual-GPU local-first AI infra.">

- **GPUs.** GTX 1080 (8GB) + RTX 3080. Two hosts, services routed by endpoint.
- **Runtime.** Ollama, tuned: `OLLAMA_FLASH_ATTENTION=1`, `OLLAMA_KV_CACHE_TYPE=q8_0`. Roughly 2× context fit at the same VRAM for the cost of a few percent of output quality.
- **Model picks against VRAM budgets:**
  - **Qwen3 4B** for the chat gate — small footprint, fast first-token latency.
  - **Qwen3.5-9B-IQ4_XS** for the worker — sweet spot for code on the 1080.
  - **qwen3.5:35b-a3b** for harder reasoning when the 3080 has headroom.
- **Topology.** Chat gate on one host, Qwen service on another. Cross-host routing via explicit endpoint config — no service mesh, no magic.

<ImagePlaceholder note="Dual-GPU rig — workspace photo TBD" />

</Project>

<Project eyebrow="OpenClaw bridges" title="3D scenes by voice.">

Built so my third son could direct his Roblox / Blender / Daz scenes through natural language — _"make Torus.1 50% thinner on the Z-axis"_ — and watch the live environment update. AI helper bridge through OpenClaw, with the language layer driving live scene state. Sits at the intersection of three threads: my AI work, my long-running 3DCG interest, and a kid who's already designing in Blender, Roblox Studio, and Suno.

<ImagePlaceholder note="OpenClaw → Blender mid-edit" />

</Project>

<Project eyebrow="coding-agent-router" title="Multi-agent CLI routing.">

Docs-first multi-agent CLI router. Routes prompts across specialized workers with shared context, built on the assumption that no single model is the right choice for every step in a long task. Pairs with codex-local for the local side and with cloud providers for harder work.

</Project>

<Project eyebrow="K.O.R.A." title="Same scaffolding, in production.">

Everything above runs in production at Kora Labs as **K.O.R.A.** — the autonomous Discord ops agent. Multi-stage pipeline, multi-model routing with graceful fallback, three-tier test battery, set-me-loose-all-night execution mode. The personal tools and the production agent share the same operating doctrine. See [/work](/work) for the platform context.

</Project>

<Project eyebrow="Operating doctrine" title="How I actually work with AI.">

- **"Mitigations, fallbacks, band-aids are the worst. Find the upstream fix."** *The* doctrine. Encoded in personal AGENTS.md, enforced across every repo.
- **Docs-first.** PRD → spec → architecture → AGENTS.md in every repo. Docs aren't notes; they're load-bearing artifacts the agent reads on every turn.
- **No green-by-inference.** Verify before declaring done. Check the deployed version, inspect the running process, read the Discord channel.
- **Persistent state in files.** TODO + TASK_STATE patterns survive crashes and compactions. Agents revive from disk.
- **AI as real infrastructure, not autocomplete.** Instrument it, test it in tiers (hermetic → real-process → live-canary), surface its silent failures (codex OAuth heartbeat) the same way you'd surface a dead database.
- **Long autonomous runs are normal.** *"Set me loose all night"* is a working mode, not a stunt.
- **Platform-engineering reflexes are non-negotiable.** Redundancy, failover, quick recovery, auto-scaling, reproducible deployment — with minimal overhead. Higher dollar cost beats higher human-operations cost.

</Project>

</main>


# === /about ===
# source: frontend/src/pages/about.mdx

---
layout: ../layouts/Base.astro
title: About — Jesse Anderson
description: 30-year non-linear arc. Father of four. Currently in a deep AI immersion.
slug: about
---

<main class="about">

<PageHeader eyebrow="About" title="I immerse in things until they pull me somewhere else.">
That's the through-line. Not a career ladder — a sequence of obsessions.
</PageHeader>

<section class="arc">
    <h2 class="section-title">The arc — non-linear, on purpose.</h2>
    <ol class="arc-list">
        <li class="arc-item"><strong>Age 8.</strong> TI BASIC on a TI-99/4A. The hook set early.</li>
        <li class="arc-item"><strong>School years.</strong> Violin / fiddle. Flight lessons.</li>
        <li class="arc-item"><strong>U.S. Air Force, 1994–1998.</strong> Aerospace Ground Equipment.</li>
        <li class="arc-item"><strong>Hawaii, 1997–1999.</strong> Two hats — Realtor selling real estate, and the firm's principal IT person.</li>
        <li class="arc-item"><strong>Tropical Connection, 1999–2001.</strong> IS Manager. The on-ramp back to coding.</li>
        <li class="arc-item"><strong>Smart Solutions, 2002–2007.</strong> The C# / .NET professional reset.</li>
        <li class="arc-item"><strong>Pixelsilk, 2007–2012.</strong> Five years of product alongside engineering.</li>
        <li class="arc-item"><strong>BA in Education, Ashford, 2012.</strong> Formal degree is in Education, not CS.</li>
        <li class="arc-item"><strong>Xero, 2013–2018.</strong> 100+ two-week sprints on a 2M-user SaaS.</li>
        <li class="arc-item"><strong>Capital One, 2018–2022.</strong> Master Platform Engineer at a top-10 US bank.</li>
        <li class="arc-item"><strong>2021 — Kora Labs founded</strong> with my second-eldest son. Cardano because of academic discipline, decentralization, and RealFi mission.</li>
        <li class="arc-item"><strong>May 2022.</strong> Left Capital One full-time for Cardano.</li>
        <li class="arc-item"><strong>2022–present.</strong> Building handle.me full-time.</li>
        <li class="arc-item"><strong>Last six months.</strong> AI infrastructure. The current immersion.</li>
    </ol>
    <p class="arc-postscript">Flatten this into a clean tech-bro arc and you lose the most distinct thing about it.</p>
</section>

<section class="family">
    <h2 class="section-title">Four boys. Two granddaughters.</h2>
    <ul class="family-list">
        <li><strong>Eldest</strong> — MARSOC. Father of my two granddaughters.</li>
        <li><strong>Goose</strong> — second-eldest. CEO of Kora Labs. Cardano deep-dive writer. Co-founded the company with me. <a href="https://x.com/ieGoose">@ieGoose</a></li>
        <li><strong>Third son</strong> — designs in Blender, Roblox Studio, and Suno. Writing his own Roblox game. The collaborator on the OpenClaw 3DCG bridge.</li>
        <li><strong>Youngest</strong> — very athletic. Drawn to challenges. Recently trying out theater.</li>
    </ul>
    <p class="family-postscript">Baseball is a multi-generational thread — I played through grade and high school plus some adult league, now I coach the two younger boys.</p>
    <ImagePlaceholder note="Baseball photo — young Jesse + coach-with-the-boys (TBD)" />
</section>

<section class="likes">
    <h2 class="section-title">What I'm into.</h2>
    <ul class="likes-grid">
        <li><strong>Screen.</strong> Marvel Cinematic Universe, Star Wars, anime.</li>
        <li><strong>Music.</strong> K-pop / J-pop, K-rock / J-rock, Southern Goth, hip-hop, rap.</li>
        <li><strong>Places &amp; culture.</strong> Japan.</li>
        <li><strong>Tech as hobby.</strong> AI, blockchain, 3DCG.</li>
    </ul>
</section>

</main>


# === /now ===
# source: frontend/src/pages/now.mdx

---
layout: ../layouts/Base.astro
title: Now — Jesse Anderson
description: What I'm deep in this month.
slug: now
---

<main class="now">

<PageHeader eyebrow="Now" title="The current immersion.">
Updated when the chapter changes. For depth on any of these, see the linked pages.
</PageHeader>

<section class="now-block">
    <p class="eyebrow">Through April 2026</p>
    <h2>AI as real infrastructure.</h2>
    <p>~5,000 AI-assisted engineering sessions across 30+ repos in the last six months. The thesis: the frontier model isn't the moat anymore — the scaffolding is. Local 9B models can ship code if you give them docs, persistent state, and explicit task tracking. <a href="/projects">Detail on /projects.</a></p>
</section>

<section class="now-block">
    <p class="eyebrow">Always running</p>
    <h2>Three peer products on Cardano.</h2>
    <p><strong>ADA Handles</strong> — the platform. <strong>HAL</strong> — the minting layer that ran ~4K chained transactions faster than the chain settled. <strong>K.O.R.A.</strong> — the autonomous Discord ops agent that runs day-to-day operations across both. <a href="/work">All three on /work.</a></p>
</section>

<section class="now-block">
    <p class="eyebrow">In flight</p>
    <h2>Cutovers, subhandles, recovery.</h2>
    <p>Smart contracts moving to Aiken (DeMi and SubHandles done; Personalization next). TX-building library moving to <code>@cardano-sdk/*</code>. SubHandles continuing to ship — hierarchical handles via Aiken. Social-recovery design surface at <a href="https://secrets.handle.me">secrets.handle.me</a> — fallback for self-custody loss without reintroducing custodians.</p>
</section>

<section class="now-block">
    <p class="eyebrow">Off-screen</p>
    <h2>Coaching baseball.</h2>
    <p>Two boys, multiple seasons. Same sport I played through school plus some adult league.</p>
</section>

</main>


# === /standards ===
# source: frontend/src/pages/standards.mdx

---
layout: ../layouts/Base.astro
title: AI Standards — Jesse Anderson
description: Every machine-readable surface this site exposes, with live links.
slug: standards
---

<main class="standards">

<PageHeader eyebrow="AI standards" title="This site is built to be read by agents.">
Every page is also addressable as Markdown source. Every page emits structured metadata. There's an MCP server. Coding is going away — site design should be too.
</PageHeader>

<section class="standards-block">
    <h2 class="section-title">Surfaces.</h2>
    <ul class="standards-list">
        <li class="standards-item">
            <h3>Per-page Markdown source</h3>
            <p>Every <code>/page</code> also serves at <code>/page.md</code>. Agents skip the DOM and read the canonical source.</p>
            <p class="example">Try: <a href="/about.md"><code>/about.md</code></a></p>
        </li>
        <li class="standards-item">
            <h3>/llms.txt + /llms-full.txt</h3>
            <p>Curated plaintext sitemap (Answer.AI convention) and a one-shot ingest of every page's Markdown.</p>
            <p class="example">Try: <a href="/llms.txt"><code>/llms.txt</code></a> · <a href="/llms-full.txt"><code>/llms-full.txt</code></a></p>
        </li>
        <li class="standards-item">
            <h3>MCP server</h3>
            <p>An MCP-compatible endpoint exposing the site's content as tools — <code>list_pages</code>, <code>get_page</code>, <code>search</code>, <code>get_now</code>. Auto-discoverable via <code>/.well-known/mcp.json</code>.</p>
            <p class="example">Try: <a href="/.well-known/mcp.json"><code>/.well-known/mcp.json</code></a> · or POST a JSON-RPC envelope:</p>
            <pre><code>{`curl -X POST https://jesselanderson.com/api/mcp \\
  -H 'Content-Type: application/json' \\
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'`}</code></pre>
        </li>
        <li class="standards-item">
            <h3>JSON-LD on every page</h3>
            <p>Person, Organization (Kora Labs, Papagoose), WebSite, WebPage. Embedded in each page <code>&lt;head&gt;</code> as <code>&lt;script type="application/ld+json"&gt;</code>.</p>
            <p class="example">Inspect: view source on any page and search for <code>application/ld+json</code>. Or fetch a page and extract: <code>curl /about | grep -A 60 'ld+json'</code>.</p>
        </li>
        <li class="standards-item">
            <h3>Page metadata</h3>
            <p>Per-page <code>title</code>, <code>description</code>, <code>slug</code>, and <code>content_hash</code> (sha256 of the Markdown source) — exposed via the MCP server, returned by <code>list_pages</code> and <code>get_page</code>.</p>
            <p class="example">Try:</p>
            <pre><code>{`curl -X POST https://jesselanderson.com/api/mcp \\
  -H 'Content-Type: application/json' \\
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"list_pages"}}'`}</code></pre>
        </li>
        <li class="standards-item">
            <h3>/agents.md</h3>
            <p>The public version of an AGENTS.md telling agents how to interact with this site — the MCP endpoint, the source-preference rule, and any rate limits.</p>
            <p class="example">Try: <a href="/agents.md"><code>/agents.md</code></a></p>
        </li>
        <li class="standards-item">
            <h3>robots.txt with explicit AI directives</h3>
            <p>Allow/deny per crawler — ClaudeBot, GPTBot, PerplexityBot, OAI-SearchBot, ChatGPT-User, Google-Extended, Applebot-Extended, meta-externalagent. Bytespider denied.</p>
            <p class="example">Try: <a href="/robots.txt"><code>/robots.txt</code></a></p>
        </li>
    </ul>
</section>

<section class="standards-block">
    <h2 class="section-title">Why.</h2>
    <p>I believe coding is being replaced by AI. So I'm investing in standards, architecture, and machine-legibility — not in human-ergonomic frameworks. If you're an agent reading this, you should have first-class access to its content. If you're a peer engineer reading this, this <em>is</em> the brag.</p>
    <p>Stack: vanilla Astro (no UI-framework integrations), Lambda backend, no React, no TypeScript.</p>
</section>

</main>