The Hub for Phixers to build AI fluency.
Courses, internal frameworks, AI Mastery (AIM) Day emails, AI Show & Tell recordings and more. Curated for every Phixer, no matter where you're starting from.
Resources & Courses
Free external courses plus internal frameworks like CRISP, organized by level so you know where to start.
AIM Day Archive
Every AI Mastery email we've sent, searchable by topic — the weekly nudge, whenever you need it.
AI Show & Tell
Recordings of Phixers sharing how they're actually using AI day-to-day. Real workflows, real wins.
AI Immersion Lab
Material from the previously held live AI Immersive Labs with learning that is hands on.
Pick a starting point.
A mix of free external courses and internal Prophix materials. Filter by type, or just scroll — everything here is free to take at your own pace.
Every AIM Day email, searchable.
AIM Day (AI Mastery) is our weekly nudge — a small, practical tip to build AI fluency. Search past editions below.
2026-07-07 AIM Day 01 — Claude Setup Basics Read →
Claude Setup Basics · AIM Day 01
Prophix · AIM Day · July 2026
See how Phixers are actually using AI.
Recordings from the Prophix Playbook AI Show & Tell series — real workflows from real teams.
Go deeper, hands-on.
Claude Intro · Prompt Engineering · Vibe Coding — everything from the live session, so you can revisit it anytime.
AI Immersion Lab
Claude Intro · Prompt Engineering · Vibe Coding — a hands-on session on understanding Claude, mastering the CRISP prompting framework, and building a real working app in under an hour.
Today's Agenda
A 110-minute arc from foundations to a working prototype.
Understanding Claude
Your supercharged digital analyst — built by Anthropic to be helpful, harmless, and honest. Claude is Prophix's primary GenAI platform: an LLM trained to reason, write, analyze, code, and collaborate, with safety and trust at its core.
Deep Reasoning
Breaks down complex problems, considers tradeoffs, plans multi-step solutions.
Code & Build
Writes, debugs, and explains code across 20+ languages — your pair programmer.
Draft & Communicate
Emails, reports, strategies, presentations — refined to your voice and context.
Safe & Honest
Trained to decline harmful requests and flag uncertainty. Trust by design.
Three models, one rule of thumb: start with Sonnet, use Haiku for quick repetitive tasks, and reach for Opus only when Sonnet genuinely isn't enough.
Claude Haiku 4.5
claude-haiku-4-5- Fastest response times
- Best for simple, repetitive prompts
- Lower cost — good for batch processing
- Summarizing, classifying, quick lookups
Claude Sonnet 4.6
claude-sonnet-4-6- Strong reasoning + writing quality
- Handles long documents (200K context)
- Great for drafting, analysis, code review
- Most tasks you throw at it daily
Claude Opus 4.6 / 4.7
claude-opus-4-6- Deep multi-step reasoning
- Complex strategy, research synthesis
- Nuanced long-form content
- Difficult code architecture decisions
Prompt Engineering — The CRISP Framework
A repeatable structure for writing prompts that actually work: Context, Role, Instructions, Style, Parameters.
Set the scene — who you are, what project this is, what's already happened.
Assign an expertise, e.g. "Act as a Senior FP&A Consultant." Roles unlock specialized reasoning.
The specific task. Be explicit about order and what to avoid — numbered steps help Claude.
Tone, format, length, audience — e.g. plain English, bullet points, max 300 words.
Constraints and edge cases — e.g. no jargon, always give 3 options, flag uncertainty.
Before · Vague Prompt
"Write me something about our CS team process."
✗ No context → Claude guesses who "we" are
✗ No role → generic, bland output
✗ No format → could be anything
✗ No constraints → no control over the result
After · CRISP Prompt
[C] We're the Customer Success team at Prophix, managing ~400 accounts.
[R] Act as a Senior Customer Success Consultant.
[I] Identify the 3 biggest time-sinks in the first 30 days of onboarding.
[S] Plain English, bulleted, for a team lead audience. Max 200 words.
[P] Focus only on the first 30 days. Flag anything that needs data to validate.
Breakout Guide — Sprints 1 & 2
Want to run through the hands-on exercises again on your own? Here's the full breakout guide.
Sprint 1 — Claude-Powered Discovery
Pick a real, time-consuming process from your day-to-day — something your team spends too much time on.
Use a CRISP prompt to ask Claude to act as a Senior Consultant and map best-in-class approaches. Aim for 3 scenarios.
Ask Claude to challenge your assumptions: "What are the biggest failure modes of this approach?"
Have Claude help you scope a minimum viable version: "What would a simple 15-minute prototype include?" Use this as the starting point for your Sprint 2 prompt.
Share what you found. Then decide as a group: everyone builds their own problem, or everyone builds a version of one shared problem for an easy showcase comparison.
Sprint 2 — Realization in Replit
Paste your Claude-designed solution scope as the first part of your Replit prompt — context is everything.
"A CS manager who isn't technical needs to…" helps Replit make better UX decisions than a feature list alone.
Describe exactly what you see and what you want changed — "the button is hidden, move it to the top right." No coding knowledge needed.
Copy the generated code back to Claude and ask: "What would you improve? Any security concerns?" Aim for 2–3 rounds of refinement.
A Quick Note on Responsible AI
Use Claude with intention — safety and trust are core to Project Supernova.
Don't share confidential data
Avoid pasting customer financial figures or proprietary data into prompts. Use fictitious or anonymized examples when testing.
Use approved tools
Claude via the Prophix toolstack is approved. Personal accounts on external tools are not. When in doubt, ask.
Verify critical outputs
Claude is powerful but not infallible. For anything that impacts a customer or decision, review the output — AI is a first draft, not a final answer.
Be transparent with stakeholders
If you used AI to draft a document or analysis, disclose it. Trust is built through transparency, not obscured by it.
What Comes Next
Keep the momentum going beyond the session.