Advanced GP
ADAP · Agentic Data Address Protocol

Every AI answer traced to its exact source. In milliseconds.

Start a pilot Google for Startups · Selected member
The problem
“I just need to know if the answers are correct.”
Every executive running AI asks some version of this. We built the answer.

We make AI answers provable.

ADAP (Agentic Data Address Protocol) is the only format that gives every line and table cell its own address, verifies every AI citation against it with precision, and opens the exact source in one click. Any AI model. Your data.

What it does

Source verification and traceability in milliseconds.

Agents use this

U001.D001.TTL1: Apple Inc.| FORM 10-Q
U001.D001.TTL2: QUARTERLY REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE ACT OF 1934
U001.D001.ORG1: Apple Inc.
U001.D001.PUB1: UNITED STATES SECURITIES AND EXCHANGE COMMISSION
U001.D001.PUBDATE1: For the quarterly period ended March 28, 2026
U001.D001.PG1.S1: PART I — FINANCIAL INFORMATION
U001.D001.PG1.S1.T1: CONDENSED CONSOLIDATED STATEMENTS OF OPERATIONS (Unaudited) (In millions)
U001.D001.PG1.S1.T1.CB: Three Months Ended — March 28,2026
U001.D001.PG1.S1.T1.CC: Three Months Ended — March 29,2025
U001.D001.PG1.S1.T1.CD: Six Months Ended — March 28,2026
U001.D001.PG1.S1.T1.CE: Six Months Ended — March 29,2025
U001.D001.PG1.S1.T1.CA.R1: Net sales:
U001.D001.PG1.S1.T1.CA.R2: Products
U001.D001.PG1.S1.T1.CB.R2: 80,208

You see this

Source Data Verification Log
Claude Opus-5.5 | Oct 16, 2026
Verified True: 2 | False: 1
✓
U001.D001.PG1.S1.T1.CB.R2: 80,208
Context: Net sales · Products · Three Months Ended March 28, 2026: 80,208
Section: Condensed Consolidated Statements of Operations
✓
U001.D010.PG1.S18.T3.CB.R7: 75.6%
Context: Gross margin percentage · Services · Three Months Ended June 27, 2026: 75.6%
Section: Gross Margin
✕
U001.D010.PG1.S12.T2.CB.R8
This address does not exist in this document. The AI cited it anyway.
For financial and professional firms

“The AI said so” isn't a source.

Agentic AI chains steps across filings, models, intranet pages and databases. An imprecise figure early on becomes a wrong decision later, and the error compounds at every step. Web pages and RAG can't check a citation or trace it to a line or cell, and the AI using them can't know whether what it cites is exact. ADAP can. Every figure an AI uses is tied to one exact address, checked by software in about a microsecond, and logged.

Only ADAP verifies.

Every citation is checked automatically and written to a Verification Log. Other formats check none.

Confirmed in one click.

The exact line or cell opens, highlighted in context.

Built around your systems.

We convert your data to ADAP and roll it out alongside the platforms your teams already run: documents today, with intranet and databases next.

100 questions on 30 SEC 10-Q filings, same AI model
155 of 155

Citations checked automatically with ADAP in under 0.2 milliseconds in total.

None checked w/ raw HTML or RAG.

0

Answers where you take the AI's word w/ ADAP.

55 for HTML and 79 for RAG.

9 seconds

For a person to confirm a typical answer with ADAP.

Versus 43 seconds with raw HTML or RAG.

Read the white paper →

Know another format that verifies every line and cell? Show us.

Start a pilot →
Data

Unblocks organizational data for agentic use.

01
Files
02
Databases
03
Web
About

Advanced GP™ is a New York startup venture.

Advanced GP AI Solutions Inc. is a Brooklyn, New York company that makes AI answers provable. We don't build AI models. We built ADAP, the only format that gives every line and table cell its own address, so every AI answer traces back to its exact source and is checked automatically. Our first product is Tasset. Advanced GP brings ADAP into financial and professional firms: we convert documents today, with intranet and databases next, connect them to the systems teams already use, and get organizations to AI answers they can put their name on.

Google for Startups

Advanced GP is a selected member of the Google for Startups program. The program provides Google Cloud, AI and Workspace credits along with engineering support.

Advanced GP uses Gemini models via Tier 3 API for development, testing and for Tasset’s in-file AI. Your file stays on your drive. Only your question and the lines the AI reads go to the model provider, under its API terms.

Team

Expert founding team from the world of finance, data and technology.

Victor Chu

Victor Chu

Founder, CEO & Head of Product
Previously Executive Director, Design Technology Innovations · JP Morgan Chase & Co.

Victor Chu wrote his first program in BASIC in fourth grade and coded until the chalk on the blackboard looked pink. He's been building at the edge of what technology can do ever since.

At JPMorgan Chase he was Executive Director of Design Technology & Innovation. His team built the 25 prototypes and the user-testing platform behind “Mobile Next,” the rebuilt Chase Mobile app that launched in 2020. In the year after launch, Chase deposits grew $305 billion and Chase became the largest U.S. bank by deposits. He also led prototyping of the Chase Digital Assistant and built the firm's first design system. Somewhere along the way, he taught wearable technology design at Parsons.

Tasset began as his capstone for MIT's executive program in Designing and Building AI Products and Services. The idea behind it was simple: if AI is going to answer for your business, it should show its work. Advanced GP is a selected member of Google for Startups, and Victor was an Element 46 Entrepreneur Fellow and a panelist at Google Cloud's AI Masterclass.

He has filed 23 patents, 11 of them issued, covering everything from smart pill bottles to a bot that chases unpaid invoices. A patent on ADAP itself is pending. He studied industrial design at North Carolina State University.

Victor lives and works in Brooklyn with his family and two dogs, Penny and Daisy, who remain unimpressed by it all.

Joe Manieri

Joe Manieri

Founding Partner, COO & CFO
Previously CFO & CAO, Fixed Income · AllianceBernstein

Joseph D. Manieri is a cofounder and a seasoned executive with extensive experience in asset management, having held several senior leadership roles at AllianceBernstein. As the Chief Financial and Administrative Officer for AllianceBernstein’s global Fixed Income division, he drove business growth and led the implementation of innovative products while overseeing financial and administrative operations. His guidance on technology investments contributed to the development of award-winning solutions and his expertise shaped the company's approach to financial management and operational efficiency.

Joe holds an MBA in Finance from Fordham University and a BS in Finance from Northeastern University. In addition to his professional pursuits, he is actively involved in community organizations and serves as a Board Member for a non-profit Catholic charity.

Harish Koya

Harish Koya

Founding Partner, Head of Engineering
Vice President, Engineering · Wells Fargo

Harish Koya is the founding engineer and is a Senior Software Architect and Java developer with over twenty years of experience delivering enterprise solutions across a range of highly regulated industries including wealth management, trading platforms, and financial services as a VP of Engineering at JP Morgan Chase & Co. and TD Ameritrade. Harish specializes in building secure, scalable, data-driven applications that enhance decision-making and operational efficiency, with strong experience in micro services architecture, real-time data integration and cloud-native deployments. Harish has joined the venture as its founding engineer to bring his passion for driving innovation and delivering robust solutions that align with business goals.

Harish holds a Graduate Certificate Program in Artificial Intelligence (AI) and Machine Learning from Columbia University and earned his MCA in Computer Science from Kakatiya University and holds AWS certifications in Machine Learning and Generative AI.