The Titan Audit

Grade your app against 28 giants

What it does

Holds every file to a bar built from 28 giants: 27 companies, from NVIDIA and Apple to Ferrari, plus NASA, each standing for three qualities. Each thing it checks turns into something you can actually test, and every problem it finds comes with how bad it is and how to fix it.

When to use it

Before you ship something you will have to look after. Or when it works and you cannot tell whether it is any good.

The prompt

This is the audit as it starts out: the whole repository, findings down to P2, and eight of the 28 giants (27 companies and NASA). On the home page you can change the mode, scope, severity and which giants apply, and the prompt rewrites itself: open the Titan console.

Audit this entire repository against the standard below.

PHASE 1 — RECON
Map the structure. Identify language, framework, database and infrastructure.
Count endpoints, models, and test files. Find the largest file. Report this
first, as a short table, before scoring anything. Read actual files. Never guess.

PHASE 2 — SCORE
Score 0–10 against each criterion below. 10 means an engineer there
would approve this specific implementation.

### NVIDIA — Visionary (foreseeing needs) · Accelerated (speeding up tasks) · Parallel (handling massive loads)
Does the shape of this still hold up when you add the next three features? Are the slow lookups saved so they only happen once? Do the things that could happen at the same time actually happen at the same time?

### Apple — Minimalist (stripping complexity) · Intuitive (zero learning curve) · Premium (high value)
Is every add-on earning its place? Could somebody new understand this in half an hour? Is there something to look at while it loads, when there is nothing there, and when it breaks?

### Amazon — Obsessive (customer/user first) · Frugal (efficient with resources) · Infinite (scalable selection/features)
Is every error message written for the person using this, not the person who built it? Does anything ask the database for an unlimited pile of rows? Could this take ten times the traffic without being rebuilt?

### Tesla — Disruptive (upending tradition) · Autonomous (self-managing) · Electric (energy/resource efficient)
Is something checking that it is alive? When a call fails, does it try again, waiting a little longer each time? Does it stop hammering a service that is already down? Are there connections or background jobs that get opened and never closed?

### Berkshire Hathaway — Patient (built for the long term) · Prudent (safe/secure) · Decentralized (modular autonomy)
Is everything arriving from outside checked before it is used? Are database questions built so that what somebody types can never become a command? Can each piece be understood, and shipped, on its own?

### Visa — Networked (connects components) · Secure (trust/validation) · Cashless (frictionless data flow)
Does every way in check who you are? Is there a limit on how fast anyone can hammer it? Is the list of sites allowed to call you an actual list, and not “anybody”? Do the keys get changed regularly?

### Oracle — Mission-critical (zero downtime) · Aggressive (competitive performance) · Database (structured data mastery)
Are the rules enforced by the database itself, and not only by the code around it? Can every change to the database be undone? Is anybody measuring which questions are slow?

### Palantir — Secretive (secure by design) · Analytical (pattern recognition) · Ontology (maps real-world to digital)
Is personal information scrambled both where it sits and while it travels? Do your records leave that information out? Does the way you organised the data match the way the real thing actually works?

PHASE 3 — FINDINGS
Report every finding down to P2. For each one give: the exact file and
line range, which criterion it fails, why it matters in one sentence, and the
concrete fix — replacement code where possible, not advice.

AUTOMATIC FINDINGS — flag every occurrence
- empty catch — P0 — catching a problem and then doing nothing hides both the failure and the cause
- password typed into the code — P0 — the moment you save it, it is in the history forever
- nothing checked on the way in — P0 — every way in is a door you left unlocked
- asking for every row — P0 — fine on a hundred rows, takes the site down at a hundred thousand
- personal details in the logs — P0 — your own records just became worth stealing
- SELECT * — P1 — you pay to move columns nobody ever reads
- any / untyped — P1 — the safety checks are switched off, but everything still looks like they are on
- console.log left in — P1 — noise that buries the one line that mattered
- no error boundary — P1 — one broken piece blanks out the entire page
- nothing checking it is alive — P1 — you find out it is down from a customer
- one file over 800 lines — P2 — nobody will read this again, including you
- copy-paste duplication — P2 — the fix now has to be made in four places
- circular imports — P2 — two files that each need the other one first; the tidy structure exists only in your head
- magic numbers — P2 — the number means something and only you know what

OUTPUT
1. The recon table
2. A scorecard: criterion | score | the single worst example
3. A composite score out of 100
4. Findings grouped by severity, worst first
5. One thing I did not ask about but need to hear

The 28 giants

The whole rubric: 28 giants (27 companies and NASA), each standing for three qualities. Fifteen of them also come with plain-English questions, which the prompt puts under their heading; for the rest, the prompt uses what the rubric says each quality means.

The names are lenses in our rubric, nothing more. None of these organizations took part in it, and none of them endorses this site or the scores it produces.

1. NVIDIA — Visionary (foreseeing needs) · Accelerated (speeding up tasks) · Parallel (handling massive loads)
2. Alphabet — Moonshot (ambitious) · Algorithmic (logically organized) · Open (accessible/extensible)
3. Apple — Minimalist (stripping complexity) · Intuitive (zero learning curve) · Premium (high value)
4. Microsoft — Ubiquitous (integrates everywhere) · Compatible (works with legacy) · Productive (focused on efficiency)
5. Amazon — Obsessive (customer/user first) · Frugal (efficient with resources) · Infinite (scalable selection/features)
6. Meta — Social (maps relationships) · Viral (maximizes engagement) · Connected (links disparate parts)
7. Broadcom — Acquisitive (strategically modular) · Efficient (maximizes margins/output) · Integrated (unified stack)
8. Tesla — Disruptive (upending tradition) · Autonomous (self-managing) · Electric (energy/resource efficient)
9. Berkshire Hathaway — Patient (built for the long term) · Prudent (safe/secure) · Decentralized (modular autonomy)
10. Eli Lilly — Scientific (rigorous logic) · Metabolic (optimizes system health) · Innovative (breakthrough approach)
11. Walmart — Low-price (computational cost efficiency) · Logistic (masterful data distribution) · Massive (scales easily)
12. JPMorgan Chase — Fortress (robust error handling) · Systemic (critical infrastructure) · Capitalist (maximizes value)
13. Visa — Networked (connects components) · Secure (trust/validation) · Cashless (frictionless data flow)
14. Exxon Mobil — Extractive (efficient data retrieval) · Disciplined (strict resource allocation) · Scale (global infrastructure)
15. Johnson & Johnson — Credo-based (user well-being first) · Diversified (balanced features) · Trusted (reliable/stable)
16. Oracle — Mission-critical (zero downtime) · Aggressive (competitive performance) · Database (structured data mastery)
17. Mastercard — Priceless (experience-focused) · Facilitating (the rails, not the friction) · Global (universal acceptance)
18. Micron — Memory (optimized storage) · Silicon (deep-level efficiency) · Cyclical (manages load spikes)
19. Costco — Membership (high loyalty/retention) · Bulk (handles volume efficiently) · Treasure-hunt (engaging discovery)
20. AMD — Resilient (error recovery) · High-performance (top-tier speed) · Competitive (better than the status quo)
21. Palantir — Secretive (secure by design) · Analytical (pattern recognition) · Ontology (maps real-world to digital)
22. AbbVie — Specialty (solves complex edge cases) · Therapeutic (fixes user pain points) · Acquisitive (absorbs best practices)
23. Bank of America — Digital (mobile/remote first) · Consumer (accessible to all) · Accessible (high availability)
24. Home Depot — DIY (empowers the user) · Helpful (clear documentation/errors) · Orange (distinct identity)
25. Netflix — Bingeworthy (keeps users engaged) · Algorithmic (smart recommendations) · Original (unique value prop)
26. Disney — Magical (delights the user) · Nostalgic (leverages familiarity) · Immersive (full experience)
27. Ferrari — Exclusive (high demand/quality) · Racing (performance-bred) · Passionate (inspires loyalty)
28. NASA — Exploratory (pushes boundaries) · Scientific (knowledge-driven) · Pioneering (first-of-kind tech)

Test log

Each prompt was pasted into a fresh private chat with memory off and graded against pass standards written before the run (all of them, dated).

The pass standard for this prompt: six criteria and an automatic-fail rule, written down on September 27, 2026, before any run. A second version for a reworded prompt was written on September 28, 2026; nothing has been graded against it yet.

Related words

• human verification gate — The one point where a named person looks at the work and takes responsibility for it going out.
• success criteria — The things you can actually watch happen that tell you the job is finished.
• verify ruthlessly — The closest thing this crowd has to a motto: use the tools hard, then check what comes back without mercy.
• vibe debt — Shortcuts and mess piling up as fast as a machine can type.