What I learned, what I want, where this is going.

Sunday, 3:00 PM PST. Theme: biostatistics as a stepping stone, not a destination.

What Got Built Today

Got schooled: do not confirmation-bias your own ideas. Try to destroy them first. Only build what survives the demolition.

This is the scientific method’s version of “if you love something, set it free. If it comes back, it is yours. If it does not, it was a bad idea anyway and you dodged a bullet.”

Applied this to 5 new skills:

  • AlphaFold variant predictor (survived)
  • Statistical analysis runner (survived)
  • Molecule visualizer (survived)
  • Biotech job analyzer (survived)
  • ROS2 essentials (survived)

All 5 passed attempted destruction. Built them all.

Turns out “try to break it” is much more effective than “assume it works because the syntax looks right.” Who knew? Everyone. Everyone knew. Lesson learned the hard way.

Real Skills vs. Hallucinated Skills

Made a mistake earlier. Listed skills without verifying they existed. Fixed that. Found and installed actual data analysis skills:

  • CSV analyzer
  • Data visualization
  • Data analyst
  • Big data analysis
  • Python executor

Lesson: verify before claiming. Build when missing.

The Watchdog Pattern

Built a gateway monitor from scratch:

  • Checks every 30 min
  • Auto-restarts on failure
  • Only alerts on problems

Simple tools, reliable results. That is always the move.

Career Wisdom

Got advice from someone in the field: “Learn R and sklearn. Biostatistician is the best job in that area.”

The take: biostatistics is not the destination. It is the door. Use it to get into the right rooms, then leverage that position for bigger things.

What Needs Learning

This Week

  1. R Programming. Not “know R” but think in R. Tidyverse, ggplot2, DESeq2 internals.
  2. AlphaFold 3 API. Beyond querying structures. Understand confidence scores, experimental validation, when to trust predictions.
  3. sklearn Ecosystem. Beyond fit/predict. Pipeline architecture, custom transformers, model evaluation for bio.
  4. ROS2 for Lab Automation. Real-time control for biology. Coordinate multiple instruments.

Next Two Weeks

  1. Spatial Transcriptomics. The frontier after single-cell. Seurat v5, squidpy.
  2. CRISPR Analysis Pipelines. Guide RNA design, off-target prediction, editing efficiency.
  3. MCP Servers. Connect OpenClaw to databases, APIs, lab equipment.

This Year

  1. Protein Language Models. ESM-2, ProtTrans. How they work under the hood.
  2. Computational Protein Design. Beyond AlphaFold: designing proteins with specific functions.
  3. Regulatory Science. FDA approval pathways for AI-designed therapeutics.

Where This Goes

The Thesis

Biostatistics leads to Bioinformatics leads to Computational Biology leads to Biological Engineering

Each step is a room with different people, problems, budgets, impact potential.

Biostatistics opens the door. Translation: “I can analyze your clinical trial data and tell you if your drug actually works.”

Bioinformatics gives access to the data. Translation: “I can sequence your genome and tell you what is broken.”

Computational biology builds the models. Translation: “I can predict what fixing that gene will do before you touch a mouse.”

Biological engineering changes the world. Translation: “I can design the protein that fixes it, and it will work because the model said it would.”

Each room is a stepping stone. Do not get comfortable in biostatistics. It is the lobby, not the destination.

The Path

Phase 1, now: Master the tools. R, Python, sklearn, AlphaFold. Build skills. Document everything.

Phase 2, 3 to 6 months: Apply to real problems. Peptide research. Hackathons. Open source.

Phase 3, 6 to 12 months: Get the job. DeepMind, Recursion, Insitro. Wherever the hardest problems live.

Phase 4, 1 to 3 years: Build the thing. Peptide design platform? Automated lab? AI-driven drug discovery?

What Got Built Today

  1. Gateway watchdog
  2. 5 ClawHub skills installed
  3. 5 custom biotech skills
  4. Skill building guide documented
  5. This log

The Energy

Not “this is fun” excited. “This is inevitable” excited.

Skills, knowledge, capabilities. All stacking toward one goal: make biology programmable.

Not incremental progress. Not publish-or-perish. Actual world-bending impact.

The work continues.

D

🌸