0:06Transcript
AI Designed a Melanoma Vaccine. The Funding Boom Is Why It Exists.
An AI designed a personalized melanoma vaccine. Why money pouring into generative AI is what funds treatments like this.
NEXUSTRADE / VIDEOS
AI trading videos
Shorts I filmed while building NexusTrade.
Every video has a transcript.
138 shorts · every one transcribed
AI videos
Start watching ↗0:06Transcript
1:38Transcript
Best AI Coding Tool in 2026? Codex Is S Tier. Cursor Fell From A to D.
My 2026 AI coding tool tier list: Codex is S tier, Claude Code is A, ZCode and Muse Code are B, and Cursor fell from A tier to D tier in six months.
3:08Transcript
Pelosi Trades Are Public PDFs. MapReduce on the Ones OCR Can't Read.
Congressional trade PDFs are public, but many are scans. I run MapReduce on them: OCR and an LLM read each page, and a reducer settles the disagreements.
1:40Transcript
AI Art Is Art
AI art is art. Here is my response to the argument about how it is created.
3:22Transcript
Why Smaller Language Models Hallucinate More. I Turned Down $475,000.
Why smaller language models hallucinate more. Senior AI engineer. I turned down $475,000. More weights and more data make a more accurate function.
5:41Transcript
MapReduce for LLMs: How Cheap Models Beat Claude Opus 5 on 100 PDFs
I teach MapReduce for LLMs with a concrete test: is Hello World in these 100 PDFs? Split the task, map it across cheap models on different views of the input, then let one reducer combine their answers.
1:26Transcript
Why Is My Backtest So Slow? GPT-6 Astra Cut Mine to 51 Seconds.
My backtests took ten minutes. GPT-6 Astra profiled the Rust engine, cited a 2014 paper, and got the run down to 51 seconds on hardware I already owned.
1:01Transcript
Ranking every AI coding tool: Claude Code, Cursor, Codex, Copilot
My tier list of five AI coding tools I use daily: Claude Code, Cursor, Copilot, OpenCode and Codex, and why Codex jumped from C tier to S tier.
0:15Transcript
Input Caching Versus Luna: My Experience
My experience with input caching and Luna: latency, cost and where caching actually helps.
2:21Transcript
Building an AI to Predict My Next Post Performance
Here's how I plan to train an AI to predict how my next post will perform: real content, features, embeddings, and the sequence of posts on my profile.
1:21Transcript
Removing the Thought Field Hurt My AI Agent Evaluation
Removing the thought field dropped my agent evaluation score from 89.9% to 85.7%. It saved some latency, with no measurable cost change. Here is the experiment.
3:05Transcript
My Latest Crazy AI Project
My latest AI project: building a model to predict how my posts will perform. Here is what I am creating and why.
2:14Transcript
How I Use Codex to Find and Apply to Software Engineering Jobs
How a software engineer uses Codex to find remote, high-paying jobs, apply directly on company sites, and land offers at Coinbase, Figma and Headway.
0:45Transcript
Claude Haiku 5.5 Cost Me 13x More Than Muse Spark
A real-world cost test of Claude Haiku 5.5 against Meta's Muse Spark on an AI trading agent's long prompts: 13x the cost, much faster, one win.
3:39Transcript
How to Pass the AI Coding Interview at Big Tech Companies
How to pass the AI coding interview: understand the codebase with a diagram, plan, review the plan out loud, and finish with test-driven development.