CorX Labs

Profile · Jamaica

Nathan Lyttle

Fourteen, from Jamaica, and the founder of CorX Labs. He has built six open-source AI models and an Android app on his own, on a single GPU, and published each one with its limits written down.


Nathan Lyttle at a glance

Nathan Lyttle is a 14-year-old developer from Jamaica and the founder of CorX Labs, an independent AI research lab. Working alone, he has built CorX1.5, a language model trained from random weights; CorX3.8-27B, Jamaica’s first large open-weight language model; NatureV1, an AI weather and hurricane model; Pluvium, which forecasts rain hour by hour for every Caribbean town; TriStream-SVS and HELIX; and Automa, an Android app for running AI agents. Every model is open source under Apache 2.0.

Name
Nathan Lyttle
Age
14 (2026)
From
Jamaica
Role
Founder, CorX Labs
Team
Works alone
Hardware
A single GPU
Models
5 open source · 4 with weights
Apps
Automa, for Android
Contact
lkk89002@gmail.com

What he has built

Every model and app Nathan Lyttle has built, with what it is and when it was announced
WorkWhat it isAnnounced
CorX1.5A 158M-parameter language model written and trained from random weights: its own architecture, tokenizer and three-stage training pipeline.Aug 2026
TriStream-SVSA 321.8M-parameter singing voice synthesis model with separate streams for pitch, timbre and breath.Aug 2026
CorX3.8-27BA 27B assistant that speaks Jamaican Patois, and Jamaica’s first large open-weight language model. A fine-tune of Qwen3.8-27B, labelled as one.Aug 2026
HELIXA sequence architecture that scales linearly with context and keeps exact long-range recall, published as helix-lm.Sep 2026
AutomaAn Android app that runs a company of AI agents from a phone, on free AI models with no key. Version 1.9.18 in November 2026.Sep 2026
NatureV1An 89M-parameter AI weather and hurricane research model whose hurricane track beats physics-only steering at every lead on held-out seasons.Oct 2026
PluviumRain forecasts for every town, hour by hour and daily to 14 days, and hurricane tracks, on the NatureV1 engine; scored for a full year in seven places against ECMWF and NOAA.Nov 2026

Alongside them: the CorX Labs benchmark index, which compares 189 AI models from 42 labs; the CorX chat; and this website.

How he got here

  1. At twelve, he wired up a chatbot on top of Gemini through its API and shaped how it behaved. It worked, but it was somebody else’s model.
  2. At thirteen, he tried fine-tuning. Slow GPUs and long runs got the better of it, and he stopped, which is why the next attempt started from random weights.
  3. At fourteen, he wrote and trained CorX1.5 from scratch. Through 2026 came TriStream-SVS, CorX3.8-27B, HELIX, Automa, NatureV1 and Pluvium.

The longer version, attempt by attempt, is on the developers page.

How he works

  • Alone, from Jamaica. Every model, the Automa app, the benchmark index and this website.
  • On one GPU. CorX1.5 was pretrained on about 3 billion tokens; NatureV1 trained in about a week.
  • In the open. Weights on Hugging Face, code on PyPI, every model under Apache 2.0.
  • With the limits written down. CorX3.8-27B is labelled a fine-tune. NatureV1’s page says its hurricane track errors are 2.5 to 4.5 times the National Hurricane Center’s, and that it is not for real decisions.

“For years I’ve shared a dream to build something great — my own AI. Today that dream came through, and I can say I made it from scratch, from the ground up.”

Nathan Lyttle, founder of CorX Labs

For writers, lists and events

Writing about young developers, student founders or AI in the Caribbean? These are the checked facts, and three bios you can use as written.

One line

Nathan Lyttle, 14, is a Jamaican developer and the founder of CorX Labs, where he has built six open-source AI models and the Automa app on his own.

Short

Nathan Lyttle is a 14-year-old developer from Jamaica and the founder of CorX Labs. Working alone on a single GPU, he has built six open-source AI models (including CorX3.8-27B, Jamaica’s first large open-weight language model, and NatureV1, an AI hurricane model) and Automa, an Android app for running AI agents.

Long

Nathan Lyttle is a 14-year-old developer from Jamaica and the founder of CorX Labs, an independent AI research lab he runs alone. He started with AI model APIs at twelve, gave up on fine-tuning at thirteen, and at fourteen wrote and trained CorX1.5, a 158-million-parameter language model, from random weights. Since then he has released CorX3.8-27B, a Jamaican Patois-speaking assistant and Jamaica’s first large open-weight language model; TriStream-SVS, a singing voice synthesis model; HELIX, a linear-scaling sequence architecture; NatureV1, an AI weather and hurricane research model; Pluvium, which forecasts rain hour by hour for every Caribbean town; and Automa, an Android app for running a company of AI agents. Every model is open source under Apache 2.0, trained on a single GPU.

Facts to check

  • Full name Nathan Lyttle; age 14 in 2026; from Jamaica.
  • Founder and only developer of CorX Labs.
  • Six open-source AI models, four with trained weights on Hugging Face (Sigmandndnns); HELIX is an architecture with no trained checkpoints, and Pluvium’s full-size network has a first training checkpoint on Hugging Face (CorxLabs), with the finished model still to come.
  • CorX1.5 is trained from random weights; CorX3.8-27B is a fine-tune of Qwen3.8-27B.
  • One app, Automa, for Android 8.0 or newer.
  • Press, events and business: lkk89002@gmail.com; on X, @corxlabs.

Questions

Who is Nathan Lyttle?

Nathan Lyttle is a 14-year-old developer from Jamaica and the founder of CorX Labs, an independent AI research lab. He is the only person who has worked on it: every model, the Automa app, the benchmark index and this website.

How old is Nathan?

Nathan is 14 (as of 2026). He started working with AI model APIs at twelve, tried fine-tuning at thirteen, and wrote his first model from scratch, CorX1.5, at fourteen.

What has Nathan built?

Six open-source AI models and one app: CorX1.5, a 158M language model trained from random weights; TriStream-SVS, a singing voice synthesis model; CorX3.8-27B, Jamaica’s first large open-weight language model; HELIX, a linear-scaling sequence architecture; NatureV1, an AI weather and hurricane research model; Pluvium, an hour-by-hour rain forecasting model for every Caribbean town; and Automa, an Android app for running a company of AI agents.

Did Nathan build his AI models from scratch?

CorX1.5 was written and trained from random weights, the architecture, the tokenizer and the training pipeline. TriStream-SVS, HELIX, NatureV1 and Pluvium are his own designs. CorX3.8-27B is a fine-tune of the open-weight Qwen3.8-27B base, and the lab says so.

Does Nathan work alone?

Yes. CorX Labs is a one-person lab. The models, their training, the documentation, the Automa app and this website were all built by Nathan, on a single GPU.

Where is Nathan from?

Jamaica. CorX Labs is an independent AI research lab and research station in Jamaica, in the Caribbean.

How can I contact Nathan?

By email at lkk89002@gmail.com, or through the contact page. Media, event and business inquiries are welcome. CorX Labs also posts updates on X as @corxlabs.

CorX Labs

Built in Jamaica, from the ground up.

Six open-source AI models and one app, all by one developer.