Partner Story · in conversation with GlyphLock. This is a written interview submitted through Submit Your Story. The claims below are the founder's own, presented as they were given to us. No payment was involved.
Carlo Earl, founder and CEO of GlyphLock. Photo: courtesy of the founder.
On a Saturday night at an Arizona venue, during an electronic music event, Carlo Earl stood near the entrance and watched the machinery of nightlife operations do what it has done for decades.
The doorman checked IDs by hand. The line grew, and he got flustered. People slipped past him. Another staff member wrote entries on paper sheets that held twenty-five names each. Once the rush started, there was no reliable count of who was inside.
"Around one in the morning, I gave my scanner to the door staff," Earl says. "I did not discover the problem later. I watched the paper process break down in real time."
Most founders describe their market research in slides. Earl describes it as a night shift, and that difference runs through everything about GlyphLock, the company he built to make sure a venue's records agree with each other and can prove what happened.
The problem: three counts, three answers
Nightlife businesses, by Earl's account, still run on paper sheets, separate terminals and systems that do not talk to each other. The door count says one thing. The register says another. The card terminal says a third. And when a payment is disputed weeks later, it can be genuinely hard to prove what happened that night.
That is not a technology gap so much as an evidence gap, and it is expensive exactly when it matters most: in a chargeback, a licensing inspection, a wage dispute, an incident report.
GlyphLock's answer is one connected system with a division of labour. NUPS runs the operational core: the door, staff, sales, the DJ booth and the nightly close. DCE records VIP agreements and payment evidence. QR Studio connects people, records and actions. Image Lab carries protected information inside images. GlyphBot assists users across the system, and the DJ tools work with music and crowd information.
"Each part has a job, but all follow the same rules for identity, permission, records and proof."
Those rules have an origin story that any developer will wince at in recognition. The platform moved four times: Firebase, then Vercel, then Replit, now Base44, and every migration forced a rebuild.
"The same things mattered each time: identity, permission, the ledger and the evidence."
What survived four rebuilds became the doctrine: record what happened, connect it to the right person and permission, mark the time, protect it from quiet changes, require human review, make the evidence easy to check later.
Five people, eleven agents, one rule
GlyphLock LLC was registered in El Mirage, Arizona in May 2025. Earl paid for it himself and did most of the early work alone. Today there are five people covering operations, engineering, security, accounting and office support.
And eleven AI agents, working on coding, testing, research, documentation, support and design.
His framing of that ratio is the cleanest we have heard from any founder running an agent-heavy operation: "They are tools, not employees. People make decisions."
The AI does real work across the platform. GlyphBot answers questions, researches, summarises and drafts. The DJ tools build playlists from mood, music style, BPM and crowd response. Other agents inspect code, test workflows, review records and explain problems.
The most ambitious use is happening right now: GlyphLock is running AI analysis over 4K camera footage from that same Saturday night event, rebuilding entries, exits, possible re-entries, timings and estimated occupancy, then comparing the result against the paper sheets and scanner records.
Earl immediately fences the claim in a way few founders bother to.
"This is being done after the event. It was not connected live to NUPS that night. The target is for camera data to become another record NUPS can compare. AI can find and explain differences, but people decide what they mean."
The camera did not check IDs, take money or write the count, which is exactly what makes it useful: it is an independent record against which the human-made records can be tested.
The list of things the AI cannot do
Ask most companies what their AI does, and you get a feature list. Ask Earl, and the first thing he offers is the opposite.
"The most important part is what our AI is not allowed to do."
The prohibition list, as he states it: AI cannot approve or deny entry. It cannot clear an age check. It cannot give someone a system role. It cannot sign or cancel a contract. It cannot add to the ledger, write a payment, or move money. It cannot change or delete evidence, or remove an audit record. It cannot accuse or punish a worker, or decide payroll. And it cannot present a test result as a real fact.
"AI can read, compare, explain and draft. People approve important actions. AI does not get final control."
These rules, he says, are written into a 59-page, 71-clause internal document he calls the Master Covenant, which applies across every GlyphLock product.
Stop and consider what that list actually covers. Entry and age checks are legal liability. Roles are security. Contracts and the ledger are money. Evidence and audit records are what a court or a card network will one day ask for. Accusing a worker is employment law. Every item on the list is a place where an AI error stops being a bug and becomes a lawsuit, and Earl has drawn the line at precisely those points.
For a two-person bar or a venue chain evaluating any AI operations product, that list works as a ready-made procurement checklist: whatever system you are being sold, ask which of these eleven things its AI is permitted to do, and why.
Money and what is running
GlyphLock is live, and the most concrete deployment is DCE, in controlled use at an Arizona venue. By Earl's account, each card swipe creates a separate agreement, a manager reviews it, and the system produces PDF and JSON evidence files for operations and payment disputes.
"This is real venue use, not a demonstration."
On pricing, Earl says creators and businesses pay $200 a month for Professional creation tools or $2,000 a month for Enterprise, with venue operators quoted per site because equipment differs. A patent application is pending on the company's authentication work.
One note from our side, in the spirit of how Earl himself talks about evidence: not everything in this account is independently checkable from the outside. The published site quotes projects individually rather than listing those subscription prices, the Master Covenant is referenced but not publicly posted, and the venue deployment is the company's own report. None of that is unusual for a company at this stage, and Earl has been notably careful to label his own claims throughout, but readers should know which parts rest on the founder's word.
Oracle, step by step, with the brakes on
The roadmap runs through Oracle's hospitality stack, and Earl narrates it with a restraint that is worth showing rather than describing.
Oracle PartnerNetwork approved GlyphLock's base enrolment on August 19. Marketplace enrolment followed on August 25. The company then connected to the Oracle OHIP Partner Sandbox and pulled roughly 250 records through read-only requests. "Nothing was changed, and the login information stayed on the server."
An August 12 test showed one sandbox request completing in 921 milliseconds, and here is what he does with his own best number: "That measures one test request, not production speed or the speed of the full platform."
The Simphony integration request is submitted and under review. "It is not approved yet." The goal, when it lands, is to bring verified sales information into NUPS and DCE, while keeping AI away from transaction writes, money movement and the ledger, the same boundary as everywhere else in the system.
There are also discussions underway around Microsoft's Independent Software Vendor programme and preparation for a second venue. Both, he says plainly, are still in progress.
The lesson from the door
Asked for the moment that mattered, Earl returns to that Saturday night, and to what he refuses to claim about it.
"I will not call that night a completed NUPS deployment because the camera, scanner, paper records and ledger were not connected. Connecting those records is the problem NUPS is meant to solve."
That sentence is the whole company in miniature. A founder watching his target problem happen live, at the door, at one in the morning, and still declining to round the story up into more than it was.
Nightlife is not a glamorous corner of the software market. It is cash-heavy, dispute-prone, staffed by people under pressure, and regulated at the door by law. Which is exactly why it is a serious test for the idea GlyphLock is built on: that the value of AI in a business like this is not autonomy, but better evidence, with people kept firmly on the decisions.
The working system is at glyphlock.io.
---
This is a Partner Story: a written interview with a company building with AI, submitted through Submit Your Story and published free of charge. Statements about the company's product, customers and results are its own. Building something with AI? Tell us about it.
