Skip to content

Commit 0ba6d0b

Browse files
manuelcorpasclaude
andcommitted
Final review fixes: Quick Path to top, decision tree, citations
- Moved Quick Path above Objectives (most actionable element first) - Replaced generic Objectives with persona-specific value propositions - Replaced "Not sure which track?" with 6-line decision tree - Added citation to 26% reproducibility stat (Garijo et al. 2013) - Reframed Databricks slot as "Scaling genomics" (neutral) - Consolidated comms to Discord as primary channel - Removed duplicate Quick Path section Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1 parent 6ba8197 commit 0ba6d0b

2 files changed

Lines changed: 31 additions & 30 deletions

File tree

docs/hackathon/index.md

Lines changed: 30 additions & 29 deletions
Original file line numberDiff line numberDiff line change
@@ -41,16 +41,28 @@ A skill is not a wrapper around someone else's code. It encodes domain judgement
4141

4242
---
4343

44-
## Objectives
44+
## Quick Path
4545

46-
By the end of this hackathon, you will have:
46+
| Step | Time | What |
47+
|------|------|------|
48+
| 1. [Setup](setup.md) | 10 min | Clone repo, install dependencies, run a demo |
49+
| 2. [Your First Skill](first-skill.md) | 20 min | Scaffold a skill, write SKILL.md, add demo data |
50+
| 3. [Add Python](add-python.md) | 20 min | Implement the skill logic with a CLI endpoint |
51+
| 4. [Test and Submit](submit.md) | 10 min | Validate, test, open a PR |
4752

48-
1. **Built a working bioinformatics skill** with a SKILL.md contract, demo data, and a runnable endpoint
49-
2. **Experienced agentic development first-hand** using coding assistants and AI agents to accelerate your work
50-
3. **Contributed to an open-source project** with a pull request to the ClawBio repository
51-
4. **Connected with a cross-disciplinary community** spanning AI engineering, genomics, proteomics, clinical diagnostics, and epidemiology across Imperial, KCL, the Crick, UCL, and industry
53+
These timings assume your environment is already set up. First-time Git or GitHub users should allow extra time; helpers will be available throughout.
54+
55+
---
5256

53-
Whether you are an AI engineer new to genomics or a genomics researcher new to agentic tools, you will leave with a concrete open-source artefact: a working skill, a documented scientific contract, and a pull request.
57+
## What You'll Walk Away With
58+
59+
- **If you are a genomics researcher**: your existing pipeline wrapped as a one-command, reproducible tool that any AI agent or colleague can run without calling you.
60+
- **If you are an AI engineer**: a working bioinformatics skill built on real public APIs, with domain decisions you can point to as evidence you understand the biology.
61+
- **If you work in proteomics**: the first proteomics skill in ClawBio. Pioneer status.
62+
- **If you work in clinical diagnostics**: a prototype skill with explicit safety rules and CPIC evidence levels, demonstrating how agentic tools can handle sensitive domain logic.
63+
- **If you work in epidemiology or public health**: a population-level analysis tool with visual output that non-specialists can act on.
64+
65+
Everyone leaves with a concrete open-source artefact: a working skill, a documented scientific contract, and a pull request.
5466

5567
---
5668

@@ -109,7 +121,7 @@ Food and drinks will be provided (pizza at 13:30).
109121
| **12:00** | **Doors open** | Arrive, get set up, connect to WiFi |
110122
| **12:30** | **Welcome** | Nathan Skene (Imperial) and Manuel Corpas (Westminster) |
111123
| **12:45** | **[Overview of the day](presentation/)** | Themes, tracks, how submissions work |
112-
| **13:15** | **Databricks** | Toz Ozturk (Databricks) -- genomics on Databricks (~15 min) |
124+
| **13:15** | **Scaling genomics** | Toz Ozturk (Databricks) -- how cloud infrastructure handles large-scale genomic workloads (~15 min) |
113125
| **13:30** | **Pizza** | Lunch break, keep hacking if you want |
114126
| **14:00** | **Guided Tutorial (Jay Moore)** | Hands-on with helpers. Follow the [Agentic Tools tutorial](agentic-tools.md) and the [ClawBio Setup](setup.md) |
115127
| **14:30** | **Build** | Work on your skill (solo or in teams). Helpers circulate. |
@@ -135,21 +147,6 @@ Both paths converge at the same goal: a working skill you can submit as a PR.
135147

136148
---
137149

138-
## Quick Path
139-
140-
| Step | Time | What |
141-
|------|------|------|
142-
| 1. [Setup](setup.md) | 10 min | Clone repo, install dependencies, run a demo |
143-
| 2. [Your First Skill](first-skill.md) | 20 min | Scaffold a skill, write SKILL.md, add demo data |
144-
| 3. [Add Python](add-python.md) | 20 min | Implement the skill logic with a CLI endpoint |
145-
| 4. [Test and Submit](submit.md) | 10 min | Validate, test, open a PR |
146-
147-
These timings assume your environment is already set up. First-time Git or GitHub users should allow extra time; helpers will be available throughout.
148-
149-
Also see the [Presentation](presentation/) for the full-screen slide deck.
150-
151-
---
152-
153150
## Choose Your Track
154151

155152
We have attendees ranging from AI agent engineers with no genomics background to researchers with 40+ years in computational biology. Pick the track that fits you. Each project is labelled as a **90-minute build** (achievable in the afternoon) or a **stretch build** (ambitious, likely a prototype).
@@ -231,9 +228,14 @@ Population-level data, outbreak analysis, and health equity.
231228

232229
## Not Sure Which Track?
233230

234-
Start with [Setup](setup.md) and run a few demos. See what clicks. Then pick a skill idea, or invent your own. The best skills come from scratching your own itch: what analysis do you do repeatedly that could be a one-command tool?
231+
- You have never written a bioinformatics pipeline: **Track A**
232+
- You work with VCF files or NGS data daily: **Track B**
233+
- Your primary data is protein or mass spec: **Track C**
234+
- You interpret genetic variants in a clinical context: **Track D**
235+
- Your work involves populations, surveillance, or disease burden: **Track E**
236+
- None of the above: **pick any 90-minute build** from any track, or invent your own
235237

236-
If you want to team up, find someone from a different track. An AI engineer paired with a genomics researcher is a powerful combination.
238+
The best skills come from scratching your own itch: what analysis do you do repeatedly that could be a one-command tool? If you want to team up, find someone from a different track. An AI engineer paired with a genomics researcher is a powerful combination.
237239

238240
---
239241

@@ -250,7 +252,6 @@ Skills will be evaluated on:
250252

251253
## Communication
252254

253-
- **On the day**: WhatsApp group (link shared at the venue) and Discord with RoboTerri
254-
- **GitHub**: [ClawBio/ClawBio Discussions](https://github.com/ClawBio/ClawBio/discussions)
255-
- **Discord**: [discord.gg/EEp4Neaz](https://discord.gg/EEp4Neaz)
256-
- **Telegram**: [ClawBio Contributors](https://t.me/ClawBioContributors)
255+
- **Primary channel**: [Discord](https://discord.gg/EEp4Neaz) -- for technical help, PR reviews, and post-event community. Monitored by maintainers.
256+
- **On the day**: WhatsApp group (link shared at the venue) for logistics
257+
- **GitHub**: [ClawBio/ClawBio Discussions](https://github.com/ClawBio/ClawBio/discussions) for longer-form questions and skill proposals

docs/hackathon/presentation/index.html

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -276,7 +276,7 @@ <h2 class="title">Biology is trapped in PDFs.<br><span class="em">AI hallucinate
276276
<div class="card">
277277
<div class="card-icon">&#128196;</div>
278278
<div class="card-heading">26% reproducibility</div>
279-
<div class="card-body">Only <strong style="color:var(--text);">1 in 4</strong> computational biology papers can be reproduced without emailing the authors. The rest? Broken links, wrong Python versions, hardcoded paths.</div>
279+
<div class="card-body">Only <strong style="color:var(--text);">1 in 4</strong> computational biology papers can be reproduced without emailing the authors. The rest? Broken links, wrong Python versions, hardcoded paths.<br><span style="font-size:0.7rem;color:var(--muted);margin-top:0.3rem;display:block;">Garijo et al., PLOS ONE 2013; Collberg &amp; Proebsting 2016</span></div>
280280
</div>
281281
<div class="card">
282282
<div class="card-icon">&#129302;</div>

0 commit comments

Comments
 (0)