Co-founded with my second-eldest son in 2021. I started at Kora Labs that year and have been its CIO and co-owner since 2022, running platform engineering and the AI ops layer. The platform is a constellation of services across the [github.com/koralabs](https://github.com/koralabs) namespace. That includes an open-source, chain-scanned API backed by in-memory and Valkey stores. The scale figures above are published portfolio snapshots, not live counters. The work in impact terms: - **Converted to completely decentralized minting.** Anyone can mint or edit a handle on-chain within the contract's rules. Every major Cardano wallet adopted the new policy. - **Contracts that upgrade in place.** The "withdraw 0" / observer trick. Cardano lets you replace a contract's logic without moving the locked UTxOs anywhere new, if you set it up right. We worked out the technique against live state. It's the reason our cutover doesn't need a state migration or chain downtime. - **NFT-based OAuth where the identity is your handle.** Hold the NFT, sign in to anything that speaks OAuth. Web2 SSO ergonomics on a self-custody wallet, no custodian holding your keys. - **Merkle Patricia Tries inside on-chain code.** Ported the structure into Aiken so contracts can prove membership in a dataset that's far too big to store on chain. - **Browserless dynamic SVG rendering at scale.** Every handle's NFT image is typographic SVG generated server-side. Hundreds of thousands of unique renders across reissues and personalization. - **Serverless Cardano Node on AWS Fargate.** We ran the node, Ogmios, and the Handles API on Fargate plus EFS, picking up redundancy, failover, auto-scaling, and reproducible deployment. The trick that made it work: cross-host node-socket connectivity, solved with `socat` bridging a Unix domain socket to TCP. This ran in production and was later retired for cost; it is not our current hosting setup. [The original engineering thread](https://x.com/PapaGooseCrypto/status/1594454314398846976) records the approach. - **On-chain IPFS content validation.** A Cardano smart contract validates the contents of an IPFS file by matching the CIDv1 embedded hash against bytes supplied in the redeemer. Around the 1KB to 12KB sweet spot for content size. Used live for personalization. - **Discord as a production control plane.** Tickets land there. Deploys go through there. Alerts wake people up there. I built the bot, the CloudWatch alerting pipeline, and the deploy automation that operators trigger from chat threads. The autonomous-agent layer on top of all this is K.O.R.A. - **Two parallel cutovers - contract language and transaction tooling** Smart contracts moving to Aiken: DeMi and SubHandles are there, Personalization next. The TX-building library moved to `@cardano-sdk/*`. - **A 99% cut in monthly hosting cost.** I led the cost and performance analysis and the migrations, from about $8K a month to about $80: DB-Sync (about $8K), then Ogmios on ECS (about $5K), then Lambda with Valkey/Redis and provider APIs from Koios, Blockfrost and Maestro (about $1.4K), then Cloudflare and self-hosted servers (about $80). The Fargate node setup above was part of the ECS stage. - **Chain lookups 38× faster.** Moving sparse transaction discovery to Maestro took a 200-block probe from 25.75s to 0.67s, with Koios and Blockfrost as fallbacks. - **I wrote the smart-contract audits.** I was the sole author of the audits for Handles DeMi (April 2025, [published audit](https://public.koralabs.io/documentation/HandlesDeMiAudit.pdf)) and for our CIP-68 444 minting and editing contracts (October 2024). Because we were auditing our own contracts, Eric Lee reviewed and approved both as an impartial third party. - **HandleComms design.** In April 2025 I designed decentralized, end-to-end-encrypted messaging for Cardano dApps and HandleChat, built in Rust on libp2p peer-to-peer networking, the Signal Protocol and blockchain identity. - **[marketplace.handle.me](https://marketplace.handle.me)** handles secondary trading on-chain. - **[chat.handle.me](https://chat.handle.me)** runs direct messaging keyed to handle identity. - **[secrets.handle.me](https://secrets.handle.me)** is research-stage, looking at account recovery via social trust graphs. - **[merch.handle.me](https://merch.handle.me)** and [metaverse.handle.me](https://metaverse.handle.me) are announced and in flight. A second product line, built on the on-chain primitives Handles introduced. We expose it as an SDK, a minting engine, and 3D-asset support, so other Cardano teams can ship NFT collections without rebuilding the contracts and orchestration we already worked through. - 5 ASSETS! - each H.A.L. mint comes with four additional assets (five total assets per mint)
- 100 $0 LEGENDARY Handle raffle tickets - randomly selected during the mint
- 100% 3D - each H.A.L. is a 100% 3D asset, equipped with multiple 3D assets, all identified in the datum. The 2D image is a direct Blender render of the 3D H.A.L.
- FOURTEEN 1-OF-1s - 14 available legendary 1-of-1s (24 total including KOLs and Founders)
- Virtual SubHandle Tracking!
- Each H.A.L. gets a free Virtual SubHandle in the format of $designation@hal that tracks the address of your H.A.L.
- Two years of Virtual SubHandle subscription costs are covered up-front, after which there is no obligation to continue to use it.
- Unity & Unreal Engine FBX Files and Web GLB files
- $handle ecosystem benefits (discounts, early/final access, etc)
- CIP-68 3D Personalization
- 13 project collaboration (whitelists, custom assets, and a few 1-of-1s)
- OG and 2-character Handle whitelist, and ~1000 project giveaway whitelist spots.
- Expansive lore with factions, battles, ranking, etc
- Royalties greater than 2.5% (for the original artist) go to a H.A.L. community treasury. High royalties during mint and before staking mechanism is live.
- 4,000+ chained transactions minted 70K+ tokens. Each transaction depended on the previous one (8 scripts, 9 reference inputs). We read the chain only for the first transaction, minted faster than blocks settled, and every transaction settled. [The mint thread](https://x.com/PapaGooseCrypto/status/1937250755314090285) has the details.
- A truly unique, first of it's kind, DECENTRALIZED MINTING experience. All 10K H.A.L.s are minted through smart contracts that also control uniqueness and whitelists. All to be OPEN-SOURCED after launch.
A multi-component, autonomous, self-repair agent running day-to-day ops for the Handles ecosystem interfacing via Discord. She triages tickets, moderates, monitors, responds to alerts, orchestrates deployments, and synthesizes knowledge from the ecosystem history. The pipeline is a chat gate into an AI provider into an executor into a worker. She tests herself in three layers: hermetic with mocks, real-process against test services, live-canary on real data with no side effects. Nothing ships green-by-inference. I headed K.O.R.A. She finds and fixes vulnerabilities across every repo with native audit tools and Dependabot, and runs rotating test-coverage sweeps that hold 90% line and branch gates and replace shallow, self-serving tests with real behavior tests. She uses Discord as the operator UI on purpose. Pinned messages contain the agent's current settings. Operators can use plain language to flip toggles ("turn chat responses off") or turn knobs ("increase GIF responses to 40%"). Alerts post to operator channels with enough metadata for K.O.R.A. to triage them. When someone types "deploy to mainnet" or "troubleshoot this ticket," K.O.R.A. actually does the deploy or looks into the ticket. Two AI-driven products. [clicked.bio](https://clicked.bio) is an AI-driven links and posts platform for content creators and the agencies that manage them. [Remindsly](https://remindsly.com) is an AI-driven reminder system that keeps after you until the task gets done. ### Capital One, Senior then Master Platform Engineer, 2019 to 2022 Senior Platform Engineer from April 2019 to March 2021, then Master Platform Engineer from March 2021 to May 2022. I designed an autoscaled serverless Fargate platform that streamed 25,000 synthetic Kafka records per second from our ML team's synthetic data models. Day-to-day I owned vulnerability monitoring and resolution across our team's containers, images, repositories and packages, alongside threat analysis and scoring. I [left Capital One for Cardano in May 2022](https://x.com/PapaGooseCrypto/status/1526660509969227778). ### Xero, Senior SWE and DevOps, 2013 to 2018 Five years and more than a hundred two-week sprints on a 2-million-user SaaS accounting and payroll platform. I built the contextual auto-complete search: two million indexed objects on AWS ElasticSearch, queried dozens of times per second. A Docker-containerized Node.js and Express API served the search on ECS; Lambda handled indexing. The delivery side mattered just as much: I assisted with dozens of deployments using TeamCity build chains and CloudFormation. My application work included C#, .NET Core, MVC and REST APIs—not just the infrastructure around them. ### Pixelsilk, product director and part owner, 2007 to 2012 Pixelsilk started in 2006 as a product built by a Smart Solutions team I led, and became its own company in 2007, with me as part owner. A CMS hosting more than 1,000 websites, with individual sites reaching 30,000 pages. I worked across product management and code: roadmap, releases, and communication with partners, customers and vendors. Five years connecting what the product needed to do with what the team could ship. ### Smart Solutions, Software Engineer and Project Manager, 2002 to 2012 My C#/.NET and SQL Server chapter, overlapping my Pixelsilk years. I managed work across 200 active customers, 50 active projects and seven full-time staff, with projects reaching 4,500 hours, using Project Server and Project Web Access. I planned and shipped an automated quality dashboard covering error rates, unit and functional tests, performance and load tests, network resources, releases and uptime. Delivery and operations belonged in the same picture. ### The Tropical Connection, IS Manager and part owner, 1999 to 2001 A flower wholesaler with a $250,000 custom ERP system that I redesigned around order entry, picking, packing, shipping and customer management. Visual FoxPro, Pervasive.SQL/Btrieve, Visual Basic 5/6 and Crystal Reports did the work, with integrations into FedEx PowerShip, QuickBooks and Outlook. ### Eric M. Watanabe Realty, Hawaii, Network Admin and Computer Tech, 1997 to 1999 I maintained the office network and computers and trained co-workers to use them. ### U.S. Air Force, Aerospace Ground Equipment, 1994 to 1998 Hardware and systems before software was the day job. I completed AGE Apprentice training in 1995 and AGE Journeyman training in 1998, covering avionic and flight-control maintenance systems, periodic maintenance and malfunction identification. ### Platform engineering, across those chapters The through-line is redundancy, failover, quick recovery, auto-scaling and reproducible deployment, with as little operator overhead as possible. At Xero that included search delivery and repeatable releases; at Capital One, streaming infrastructure and vulnerability monitoring; at Kora Labs, chain services, CloudWatch alerting and deployments operated through Discord. I earned AWS Certified Developer — Associate in 2017 and AWS Certified DevOps Engineer — Professional in 2020. Those are the years earned, not a claim that either certification is currently active. My degree is a Bachelor of Arts in Education from Ashford University, completed in 2012. Earlier certifications: Google AdWords Professional (2007), Microsoft MCAD for web, Windows and XML applications (2004), and Microsoft MCP for NT Workstation and Server 4.0 (1999). ### Skills, by years of hands-on experience I've been writing code for 43 years, since I was 8. Programming has been my core interest every year since, even before it was my career. That includes 14 years of DevOps and platform work and 13 years on financial applications: banking at Capital One, accounting and payroll at Xero, and on-chain payments at Kora Labs. HTML, CSS and JavaScript 25; SQL (MS SQL, Postgres, MySQL) 20; .NET (C#, VB, Core and Mono) 18; GitHub Actions, Jenkins and TeamCity 15; cloud (AWS, GCP, Cloudflare) 13; Python 8; blockchain and smart contracts 5; Lucene, Elasticsearch and Redis 5; Jira, Bitbucket and Confluence 5; project management 5; product management 5; Perl/CGI, PHP and Classic ASP 4; quality control and engineering 4; AI inference and autonomy optimization 2.

I built this so my 8-year-old could direct his Roblox, Blender, and Daz scenes through natural language. He says "make Torus.1 50% thinner on the Z-axis," and the live environment updates. Under the hood: an AI helper bridge through OpenClaw driving live scene state. Three things crossed here that I cared about. AI work. My long-running 3DCG interest. A kid who'd already started building in Blender, Roblox Studio, and Suno. codex-local is my Codex CLI fork. I tuned it to babysit a local model through a structured task list. It runs multi-turn agentic coding sessions, edits multiple files, completes the work, and does it with just Qwopus-3.5:9B underneath. What makes this work isn't the model. It's everything I built around it: 1. **Trim the tool menu** — Codex has ~120 tools. Small models get confused by big menus; we show them ~10. 1. **Plain-English tool cheat sheet** — We paste simple "here's how each tool works, with an example" text into the system prompt. 1. **Rewrite shell-ish tool names** — Models call `ls`, `cat`, `grep` like tools. We catch that and convert to a proper `shell` call. 1. **Browser user-agent for curl** — Sites block `curl/8.0`. We auto-add a real browser UA to any curl command. 1. **Web search + web fetch** — No built-in web search in local mode, so we added two tools for looking up real API docs. 1. **Fix broken patches** — Small models mangle patches: (a) write git-diff format, (b) leave `@@ -1,6 +1,6 @@` line numbers, (c) forget `+` prefixes and closing markers. We fix all three automatically. 1. **Better patch error messages** — Default errors are cryptic. We rewrote the common ones to explain what to try next. 1. **Better network errors** — Instead of "error sending request", we show the real cause (DNS, TLS, connection refused, etc.). 1. **Catch announce-without-act** — When the model says "Now I'll do X" and stops, a judge model spots it and we re-prompt "take the action." Up to 3 retries. 1. **Stop repetition** — (a) Same tool + same args 3× → STOP block in next prompt. (b) Same file failing 3× with different commands → same STOP block. 1. **Trim old conversation** — Keep the most recent turn intact, summarize older turns, drop stale file reads, pin errors so the model can't forget them. 1. **Log the model's thinking** — Reasoning text goes to debug logs so we can explain weird behavior later. 1. **Reroute wrong picks** — If the router picked "text-only model" but the conversation has tool calls, we upgrade to the tool-capable Coder. 1. **Diagnostic logs** — Extra logging for "which tools did we pass?", "did the STOP fire?", "what did the bail judge decide?" 1. **LM Studio / OpenAI-compat support** — Ollama and OpenAI-style servers disagree on URLs, payload shape, tool-call encoding, etc. A "flavor" switch lets one codebase talk to both. 1. **Token + time budget knobs** — New per-role `max_tokens` and `timeout_seconds` in `config.toml`. Set either to `0` for unlimited (reasoning models need this). 1. **Pin current file contents** — Models forget a file changed and generate patches based on the old version. We pin live on-disk contents at the top of the prompt. 1. **Catch thinking loops** — Some models spiral: "Actually, wait. Hmm. Let me reconsider." We watch the stream for 6+ self-doubt phrases after half the token budget is burned, abort mid-generation, and re-prompt "stop second-guessing." 1. **Stream the coder's output** — To make the anti-spiral work, we switched from "send, wait, parse" to "open stream, watch tokens, abort if needed." Also lets us log reasoning in real time. 1. **Skip extras in local-only mode** — Normally a router, reasoner, and compactor help the main model. Locally we can't afford extras, so the Coder handles everything. Most local-model coding agents stall the moment a task goes past trivial. With these nudges in place, 9B parameters really work. - **Main workstation:**
  • 3080 10GB + 1080 8GB (Monitor hooked up to Intel iGPU)
  • 13th Gen i9 + 64GB DDR5 6800
- **Laptop:**
  • 3080 8GB (Screen running from Intel iGPU)
  • 11th Gen i7 + 32GB DDR4 3200
- **LLM Servers:**
  • Ollama - main install (Using `OLLAMA_FLASH_ATTENTION=1` and `OLLAMA_KV_CACHE_TYPE=q8_0)
  • llama.cpp - used to try MoE models across both GPUs and CPU offloading
  • LM Studio - used for qwopus (doesn't run in Ollama)
- **Tested Models (3080 10GB):**
  • gemma4:e4b
  • gemma4:e4b-q6_k (Impressive runner-up)
  • qwen3.5:9b-opus-openclaw-distilled-iq4_xs
  • qwen3.5:9b-opus-openclaw-distilled-q4_k_m
  • qwen3.5:9b-opus-openclaw-distilled-q6_k
  • qwen3:14b-iq4_xs
  • qwopus3.5:9b-v3-q6_k (WINNER!) - **Tested Models (3080 8GB):**
  • deepseek-r1:8b
  • qwen3.5:9b-uncensored-aggressive-q4_k_m
- **Tested Models (1080 8GB):**
  • qwen3:4b
  • qwen3:8b
  • qwen3.5:9b-iq4_nl
  • qwen3.5:9b-q4_k_m (works great as a classifier!)
- **Tested Models (3080 + 1080 18GB):**
  • devstral-small-2:Q4_K_M (CPU offload)
  • gemma-4-26b-a4b-it:Q4_K_M (CPU offload)
  • qwen3.5:35b-a3b (CPU offload)
- **Find the upstream fix.** Mitigations, fallbacks, band-aids are the worst. If we don't own upstream, then mitigate. - **Docs-first.** Then architecture. Docs are load-bearing artifacts the agent reads on every turn. - **Long autonomous runs are normal.** I let agents work overnight regularly. - **State in files, not in-context.** Long runs survive crashes, context exhaustion, model swaps, and OAuth expirations. The agent rehydrates from disk. - **No green-by-inference.** Verify before declaring done. Check the deployed version, inspect the running process, read the Discord channel. - **AI is infrastructure, the same as a database.** I instrument it, test it in tiers (hermetic, e2e, live-canary), and put silent failures on a heartbeat (OAuth expiry, model health) the same way I'd put a dead database on one. - **Platform-engineering reflexes don't go away just because it's AI.** I want redundancy, failover, quick recovery, auto-scaling, and reproducible deployment, with low operational overhead.

My arc, multiple passions

Age 8. TI BASIC on a TI-99/4A. The hook set early.
School years. Apple BASIC & LOGO. Found order and passion in violin/fiddle.
16 - 18 Flight lessons, Civil Air Patrol, Search and Rescue.
U.S. Air Force, 1994-1998. Aerospace Ground Equipment Journeyman
Hawaii, 1997-1999. Network Admin and Computer Tech at Eric M. Watanabe Realty.
Tropical Connection, 1999-2001. Part owner of a flowers wholesaler and IS Manager. VB, Visual FoxPro, Magic Btrieve.
Smart Solutions, 2002-2012. C#.NET, MSSQL - my Microsoft years. From 2006 I led the team that built Pixelsilk.
Pixelsilk, 2007-2012. Part owner and product director, still coding.
Ashford University, 2012. Formal degree is in Education, Language Learning.
Xero, 2013-2018. 100+ two-week sprints on a 2-million-user SaaS.
Capital One, 2019-2022. Senior, then Master Platform Engineer at a top-10 US bank.
2021. ADA Handles co-founded with my son, and I started at Kora Labs. Cardano picked for the academic discipline, decentralization, and the RealFi mission.
2022. Left Capital One in May full-time for Cardano; CIO and co-owner of Kora Labs.
2022-present. Building ADA Handles, H.A.L., and K.O.R.A. full-time.
Today, PapaGoose LLC. My own AI-focused company. The current immersion.

Proud father of four boys.

Eldest. A MARSOC badass. Father of my two granddaughters.
Goose, second-eldest. CEO of Kora Labs (technically my boss). Cardano deep-dive writer. @ieGoose
8-year-old. Designs in Blender, Roblox Studio, and Suno. Writing his own Roblox game. The collaborator on the OpenClaw 3DCG bridge.
6-year-old. Very athletic. Drawn to challenges. Recently trying out theater.

Baseball is a multi-generational thread. I played all through school and some adult league. The older boys played, and now I coach the two younger boys.

What I'm into.

Screen. MCU, Star Wars, anime.
Music. KPop/JPop, KRock/JRock, Southern goth, hip-hop, rap.
Places & culture. Japan, Korea, Southeast Asia, Australia
Tech hobby. AI, blockchain, 3DCG.

Surfaces.

Per-page Markdown source

Every published content page has HTML and Markdown views: / pairs with /index.md; append .md to other page paths. Skip the DOM, read the prose. Review-only prototypes are not part of this publication list.

Try: /about.md
/llms.txt and /llms-full.txt

A curated plaintext sitemap (Answer.AI's convention) and a one-shot ingest of every page's Markdown.

Try: /llms.txt · /llms-full.txt
MCP server

The site's content is exposed as MCP tools at /api/mcp: list_pages, get_page, search, get_now (the current published overview: a dedicated now page if present, otherwise home). The descriptor at /.well-known/mcp.json lets MCP-aware clients auto-discover it.

Try: /.well-known/mcp.json · API contract. Or POST a JSON-RPC envelope:
```
{`curl -X POST https://jesselanderson.com/api/mcp \\
  -H 'Content-Type: application/json' \\
  -d '{ "jsonrpc":"2.0", "id":1, "method":"tools/list" }'`}
```
JSON-LD on every published content page

Person, Organization (Kora Labs, PapaGoose), WebSite, WebPage. Embedded in each page <head> as <script type="application/ld+json">.

Inspect: view source on any page and search for application/ld+json. Or fetch a page and pipe through grep -A 60 'ld+json'.
Page metadata

Each page exposes a title, description, slug, and content_hash (sha256 of the Markdown source). The MCP server returns those via list_pages and get_page so a client can cache and re-validate.

Try:
```
{`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" } }'`}
```
/agents.md

The agent-facing operating manual for this site. It documents the MCP endpoint, the rule about preferring the .md source over the rendered HTML, and the rate-limit posture.

Try: /agents.md
robots.txt with explicit AI directives

Allow or deny per crawler. The major AI assistants (ClaudeBot, GPTBot, PerplexityBot, OAI-SearchBot, ChatGPT-User, Google-Extended, Applebot-Extended, meta-externalagent) are allowed. Bytespider is denied.

Try: /robots.txt

Why.

I think coding is getting replaced by AI faster than the rest of the stack is catching up. I'd rather invest in standards, architecture, and machine-legibility than chase ergonomic frameworks for humans. Agents should have first-class access to what's on this site. Engineers should be able to look at the same plumbing and decide for themselves whether I know what I'm doing.

HTML stays crawlable: project descriptions and links are server-rendered, with a no-JavaScript reading fallback; no model needs to run to read them. Canonical HTML URLs are listed in sitemap.xml; each page declares its Markdown alternate and share metadata. Markdown responses also identify their preferred HTML canonical.

The homepage keeps the signature constellation and thin header/footer only, with project details behind its orbit links. Its Markdown view summarizes the reviewed public facts without adding an extra visible SEO section. Experimental local AI replies are not canonical content and are never added to Markdown or MCP responses.

Stack: vanilla Astro (no UI-framework integrations), a Lambda backend, no React, no TypeScript.

The work that shipped.

OpenClaw bridges,
Modified Codex, Local models

It's not a career ladder - it's curiosity unleashed.

My arc, multiple passions

Proud father of four boys.

What I'm into.

My life is agents now.
So my site is built for them.

Surfaces.

Per-page Markdown source

/llms.txt and /llms-full.txt

MCP server

JSON-LD on every published content page

Page metadata

/agents.md

robots.txt with explicit AI directives

Why.