LESS WASTE. MORE INTELLIGENCE. GREATER HUMAN BENEFIT.
Evidence before
confidence.
Messy exports should not force you to guess. OBSIDIAN builds small, inspectable data tools that show what needs attention — and leave important decisions with you.
Early-stage, founder-operated project. Demonstrations below are synthetic, not customer results. No sales or investment performance is claimed.
01 / ORIGINAL02 / CHECK03 / EVIDENCE04 / YOUR DECISION
A REAL EXECUTION. FICTIONAL DATA.
Three kinds of repetition.
Three different decisions.
Our existing read-only audit checks a CSV export, separates exact repeats from formatting-only matches and conflicting keys, then reports the affected spreadsheet rows. It does not merge or delete your records.
Before / an ambiguous export
| Rows | Example | Problem |
|---|---|---|
| 2–3 | 001 / Cedar mug / A | Exact repeat |
| 4–5 | Linen bag / linen bag | Case-only match |
| 6–7 | Key 003; Cotton tote / Cotton tote XL | Conflicting details |
After / a reviewable exception list
- Exact duplicateRows 2, 3 — identical in every column.
- Near duplicateRows 4, 5 — case/spacing comparison.
- Key conflictRows 6, 7 — same ID; product differs.
No “best” record invented. Every original remains intact.
How this was verified
Expected group types and row references were specified before execution. The verifier compares exact output values, checks source bytes and SHA-256, and runs clean-input and malformed-input cases. This proves this bounded workflow, not universal correctness.
SYNTHETIC DEMONSTRATION - NOT CUSTOMER RESULTS
Reconciling invoices, payouts and bank deposits.
A bounded, file-level reconciliation audit. Synthetic exports only; no real customer data.
Inputs (fictional)
Three small CSV exports: 5 invoices, 9 processor transactions, 4 bank deposits.
Output
Matched transactions, processor fees, partial payments, refunds, duplicates, missing deposits, timing differences, batched payouts, and ambiguous matches - each with source-row references.
Ambiguous example
One bank deposit of $100.00 could belong to either an invoice payment or a duplicate processor transaction. It is flagged AMBIGUOUS and left for human review; it is never auto-matched.
File integrity
SHA-256 fingerprints are recorded for each input file. They prove the files were not changed after the run. They do not prove the transactions themselves are true.
Administrative data matching only. Not accounting, tax or financial advice. Synthetic demonstration - not customer results.
WATCH THE WORKFLOW
From an export to an informed decision.
Silent visual walkthrough; captions and full text carry all information. No simulated customer activity.
Read the complete video transcript
1. OBSIDIAN: evidence before confidence. A bounded synthetic demonstration, not customer work.
2. The problem: exact repeats, formatting-only matches and conflicting records need different decisions.
3. Inspect: the existing Python audit reads six fictional records and records the original fingerprint.
4. Report: rows 2–3 are exact repeats, 4–5 are case-only matches, and 6–7 conflict on an agreed ID.
5. Verify: independently specified groups and row references match. Source bytes remain unchanged.
6. Decide: review the evidence before changing a record. Discuss scope before payment; no live-system or import guarantee.
SMALL SCOPE. EXPLICIT LIMITS.
Start with the question, not a payment.
These are separate existing offers. Confirm scope and an approved private transfer method before payment. Confidential or identifiable customer data is not accepted until privacy, transfer and retention safeguards are verified and approved. This showcase does not establish intake readiness.
CSV Cleaning Starter
One CSV, up to 50 data rows. A cleaned copy, change log and exceptions report.
No multi-sheet workbooks, CRM imports or email verification.
Duplicate Detection & Audit
One CSV, up to 2,000 data rows. Read-only groups, row references and source fingerprint.
No merging, deletion, fuzzy name matching or CRM changes. Existing duplicate-audit contract unchanged.
Spreadsheet Quality Audit
One comma-delimited UTF-8 CSV: up to 100 data rows, 30 columns, 1 MiB. Read-only findings and column profiles.
No corrected file, accounting conclusions, macros or import guarantee.
Delivery date is agreed after scope and readiness checks. Published package terms control any eventual order; no new turnaround or refund guarantee is made here. Checkout availability and payout readiness require confirmation.
CUSTOMERS / PILOT PARTNERS / ADVISORS
Bring a problem.
Keep your data private.
Email OBSIDIAN about your project without a Fiverr account, or use the existing Fiverr contact option. The email link opens your email application with a short template; this website does not submit or send messages. You review and send from your own email application. If no email application is configured, use the displayed address in your existing email service. The optional local composer below prepares a Fiverr draft only; it sends nothing, stores nothing and does not transfer the draft automatically.
Partnerships: ask about a bounded export-review pilot. No white-label deployment, integration expertise or reseller arrangement is assumed. Advisors: the presentation explains our present proof and long-term hypothesis.
Email OBSIDIAN about my project
Opens your email application, not a website submission. No message is sent until you send it yourself. Email application unavailable? Use dustyknucks666@gmail.com in your existing email service. Opening the link does not prove delivery.
Do not send confidential information or attach files before agreeing on scope and a secure transfer method. For Fiverr-originated inquiries, keep communication and payment on Fiverr.
Direct external link, not an automated form. Login and the native message option may be required. The CA$ services above and Fiverr packages are separate offers: confirm the applicable scope, currency and platform before ordering. No cross-platform payment is requested.
LONG-TERM VISION / CONCEPT, NOT A PRODUCT CLAIM
More useful intelligence.
More human control.
We aim to build AI-assisted tools that make decisions easier to inspect, correct and trust. Today the evidence is a bounded set of data workflows and synthetic tests — not general intelligence, autonomous business operations or a proven investment return.
We want to earn the next step through useful work, consent and verifiable results. Less waste. More intelligence. Greater human benefit.