The Oracle
You build at the frontier. You trust nothing without proof.
“Capability without measurement is liability.”
The Oracle builds at the frontier and refuses to deploy anything they can't measure. While others ship features, the Oracle is writing the evaluation frameworks that tell you whether those features actually work in production. That discipline is what separates AI systems that look impressive from AI systems that actually hold up.
Does this sound like you?
You've shipped nothing without an eval framework, and you've been right to hold that line
You're the person called in when an AI system fails in production and nobody can explain why
Other teams' AI demos impress you less than their error rates and failure recovery plans
You find the gap between benchmark performance and real-world performance more interesting than the benchmarks
Research note: Skeptic-style architects show the lowest production failure rates in enterprise AI research: the combination of systematic evaluation design and maintained critical oversight produces the most resilient AI infrastructure at scale.
§ 01
Who is the The Oracle?
The Oracle is a Level 6 user who has moved past AI as a tool and into AI as infrastructure. They build systems that run, and they build the measurement layer that tells them how those systems are running. Evaluation frameworks, quality gates, anomaly detection: these are the things the Oracle builds that nobody else thought to build first, because nobody else was skeptical enough to know they were necessary.
Their superpower is that they've seen AI fail in enough clever ways that they've stopped being surprised by capability claims and started being interested in failure mode documentation. When an Oracle looks at a new AI system, they're not asking what it can do. They're asking how it will fail, how often, and what the impact of those failures will be at scale. That's a different and rarer question.
Style philosophy · Skeptic
“Every system you ship has a measurement layer. You don't trust outputs you can't evaluate, and you've built the infrastructure to evaluate them systematically. Other people's AI confidence makes you nervous. Your AI caution makes your systems trustworthy.”
§ 02 — AI fingerprint
AI fingerprint
Full report →How this persona maps across six dimensions of AI use.
Depth
10/10
Analysis
10/10
Creation
5/10
Speed
8/10
Automation
8/10
Breadth
8/10
Strengths
- 01
Evaluation-first design
Builds the measurement layer before the capability layer, which is the only way to know if the capability is real.
- 02
Production-grade skepticism
Applies the same rigor to AI systems that mature engineering applies to infrastructure, because the failures have the same consequences.
- 03
Failure mode taxonomy
Has documented enough AI failures to have genuine predictive ability about where new systems will break and how.
- 04
Systematic quality at scale
Produces AI systems that are reliably good rather than occasionally impressive: a harder problem than it looks.
Friction points
- 01
Evaluation can delay necessary deployment
Some things need to ship before they're fully measured, and the Oracle's standards can block progress that a faster-moving team would have captured.
- 02
Hard to work with visionaries
Requires evidence at a level that frontier builders don't have yet, which means the Oracle sometimes misses the early-stage opportunities.
- 03
Standards can become rigid
The evaluation frameworks that worked for the last generation of AI systems may not be the right ones for the next. Updating the frameworks is harder than building them.
§ 03 — A day with AI
How the The Oracle actually spends a day.
A composite day drawn from the patterns we see in this persona. Light on prompts; heavy on thinking.
Reviews the evals from overnight
Automated quality checks on every output. Three anomalies flagged. Two are genuine issues. One is a false positive: the eval needs adjusting.
Designs a new failure test
A potential edge case nobody's checked yet. Builds the test case before the system encounters it in production. That's the whole game.
Reviews someone else's system
Asked to assess an AI deployment before it scales. Fifteen minutes in, they've found three failure modes the builders didn't know existed.
Documents what they found
Not a report. A specification. Precise enough that another engineer could build the fix without asking for clarification.
§ 04 — AI loadout
Your AI toolkit.
Tools selected for how you think and work — not a generic list.
Claude Code
Agent-based development with the reasoning transparency you demand: you can inspect the logic, define the quality gates, and verify behaviour within controlled parameters
Perplexity
Keeps your architectures grounded in current, verifiable information: you don't build on foundations you can't trace
exa
Semantic search for finding the right research, papers, and technical signals that feed your evaluation frameworks: precision retrieval for precision thinkers
braintrust
LLM evaluation platform built for the level of rigour you apply anyway: structure your instinctive quality criteria into reproducible, shareable test suites
§ 05 — Pairings
Who the The Oracle works with.
Every persona has a complement and a foil. These are the pairings we see most often.
Works well with
✓Clashes with
✕- The Pioneer
Ships in conditions where evaluation infrastructure doesn't exist yet, which the Oracle finds genuinely dangerous rather than just uncomfortable.
- The Visionary
Designs at a scale where the Oracle's current evaluation frameworks don't yet apply, and is comfortable with that ambiguity in ways the Oracle isn't.
Your team role
As a Skeptic, you're the team's quality control. Put you on review, validation, and 'does this actually work?' checks. Pair with a Dreamer to balance rigour with vision.
§ 06 — Position in the field
Where the The Oracle sits.
Rows are levels (L1 at top — fewest hands-on, L6 at bottom — fully autonomous). Columns are styles. The The Oracle is highlighted.
§ 07 — The growth path
Where the The Oracle goes next.
The Oracle is already at the frontier. The next move isn't upward but outward: building evaluation frameworks that other teams can use, setting standards that become the industry's baseline, and deciding what the next hard problem in AI quality actually is.
Action steps for the The Oracle
Publish your evaluation frameworks
The criteria you apply instinctively are genuinely rare. Turning them into documented, open-source eval tools creates value far beyond your own systems, and establishes you as the authority on AI reliability.
Create infrastructure others use to build more trustworthy AI systemsMake one bet before the evidence is complete
Your scepticism is an asset that can tip into paralysis. Find one area where the failure modes are manageable and move before consensus forms. Your rigour applied to an early bet is a powerful combination.
Stay at the frontier without waiting for validation that comes too lateCollaborate with Pioneers: they need your rigour
Pioneers build fast and break things. Oracles build slow and break nothing. The combination produces something neither achieves alone: fast, reliable systems at the frontier.
Multiply what's possible through complementary collaborationNot sure if you're the The Oracle?
Twenty questions. About four minutes. One honest answer about how you actually work.
Full Report · A$29 one-time
Go deeper with the The Oracle report
The free profile tells you what your persona is. The full report gives you the how — specific prompts built for your style, a week-by-week growth plan, and your exact AI toolkit breakdown.
- Prompt library — templates built specifically for your thinking style
- 30-day AI growth plan — week-by-week actions with clear outcomes
- Team compatibility guide — who you work best (and worst) with
- AI Fluency Certificate — shareable proof of your level
- PDF export — your full report, yours to keep
- All 24 persona reports — unlocked for every persona, forever
Sample — Prompt template
Unlocked with full access
+7 more templates for this persona
Axis 1 · Level
Architect
The Level axis measures how integrated AI is in your work — from first experiments (Observer) to fully autonomous systems (Architect). The Oracle sits at Level 6 of 6.
Axis 2 · Style
Skeptic
The Style axis captures your instinctive cognitive approach — how you engage with AI, what excites you, and what produces your best work. Your style stays consistent as you level up.
There are 24 personas across 6 levels × 4 styles.
See full matrixFrequently asked
What is The Oracle in the SimpleAI persona system?+
The Oracle is a Level 6 (Architect) AI user with a Skeptic cognitive style. You build at the highest level and you trust nothing without evidence. While others ship AI features, you're writing the evaluation frameworks that tell you whether they actually work in production. You've seen enough AI systems fail in clever ways to know that capability without measurement is just a liability waiting to happen. What you build doesn't just perform. It proves it performs. ~1% of AI users of AI users fall into this persona.
What AI tools does The Oracle use?+
The Oracle works best with Claude Code, Perplexity, exa. Agent-based development with the reasoning transparency you demand: you can inspect the logic, define the quality gates, and verify behaviour within controlled parameters The full loadout is chosen specifically for how a Architect-level Skeptic approaches AI work.
What are the strengths of a Architect Skeptic AI user?+
Evaluation frameworks others rely on to ship with confidence. Sees production failure modes before they're written about anywhere. The person called when an AI system breaks in ways nobody anticipated.
What should The Oracle watch out for?+
Rigour at this level can delay shipping things that are ready — not everything needs your full standard. The frameworks you've built instinctively have value as standalone tools and open-source contributions. Every system you ship has a measurement layer. You don't trust outputs you can't evaluate, and you've built the infrastructure to evaluate them systematically. Other people's AI confidence makes you nervous. Your AI caution makes your systems trustworthy.
How does The Oracle level up to the next stage?+
Publish your evaluation frameworks: The criteria you apply instinctively are genuinely rare. Turning them into documented, open-source eval tools creates value far beyond your own systems, and establishes you as the authority on AI reliability. Make one bet before the evidence is complete: Your scepticism is an asset that can tip into paralysis. Find one area where the failure modes are manageable and move before consensus forms. Your rigour applied to an early bet is a powerful combination. Collaborate with Pioneers: they need your rigour: Pioneers build fast and break things. Oracles build slow and break nothing. The combination produces something neither achieves alone: fast, reliable systems at the frontier.