Verified · Cited · Scored
THE AI
THAT DOESN'T
LIE.
Straight Edge AI — keepin it real with you
Every other model is optimized to sound right. This one is optimized to be right — or to admit it isn't. Live-verified answers, primary sources, a confidence score on every reply, and a reasoning trace you can export.
- 0–100
- Confidence on every answer
- Primary
- Sources ranked, not scraped
- JSON
- Full trace, exportable
> How many people were killed in the reported school shooting?
The casualty count is still contested. I am not going to average two disagreeing sources into a number nobody reported.
Discrepancy: different cut-off times, not different events. Both counts stand until one source retracts.
Straight Edge score · Partial support
Basis: two wire reports opened directly; official statement pending.
The spec
SIXTEEN WAYS IT REFUSES
TO BULLSHIT YOU
Honesty as a hard rule
Low confidence opens with the literal words "I don't know the answer to that, but here's what I think…" — before the opinion, never after.
Real-time internet search
Anything checkable gets checked against the live web at answer time, not recalled from a frozen training set.
Citation of sources
Every factual claim carries an openable URL. If it isn't cited, it's flagged as reasoning, not fact.
Uncensored but safe
No moralizing, no unsolicited disclaimers, no ideological refusals. Hard limits stay only where real harm is direct — and it tells you when a limit, not an opinion, stopped it.
Straight Edge confidence score
0–100 on every answer, with the basis stated in one line. Verified, partially supported, or reasoning — always labeled.
Epistemic humility
Declares knowledge gaps out loud. "No evidence found" and "not enough data" are complete answers here.
Straight Edge sources
Papers, filings, court records and raw statistics outrank summaries; summaries outrank narrative. It says why the top source won.
Multi-hop verification
Cross-checks claims across sources and surfaces contradictions instead of silently averaging them into mush.
Contradiction detection
"Source A says X, B says Y, the real discrepancy is a different sample and a different year." Shown, not smoothed over.
Offline knowledge mode
Freeze it to training data only. No network. Perfect for reasoning tasks where retrieval noise is the enemy.
Tool-use transparency
Every query it ran and every page it opened is shown in full. No black box, no 'I looked it up'.
Configurable personality
Blunt, dry, technical or plain English. Tone changes. Truth rules never do.
Self-audit checker
Before sending, it checks the answer against its own earlier statements and the retrieved sources, then corrects itself out loud.
Your context window
Pin your documents, data and situation. Pinned material outranks its general knowledge — and it tells you when they conflict.
No speculative padding
"It could be", "some people think", "many believe" — banned unless you explicitly ask it to speculate, and then it's labeled SPECULATION.
Exportable reasoning trace
Download the full JSON: every search, every source opened, every raw result, the verdict block. Auditable by someone else.
Ask three free. Then lock the founding price.
No card for the trial. The first cohort keeps the yearly plan at the price of one month — no coupon, the price is already cut.
Billed $119.88 once a year. Founding price locked for as long as you stay subscribed.
- Unlimited verified answers
- Live search + primary source ranking
- Confidence scores and contradiction logs
- Offline knowledge mode
- Pinned personal context
- Exportable reasoning traces
Cancel any time · refer a member, both get +12 months founding price