DeckEdit

Agentic Engineering

Beyond Vibe Coding. DeckEdit is built with Agentic Engineering, the discipline Andrej Karpathy named in February 2026 for orchestrating AI coding agents under expert human direction. DeckEdit is the existence proof that disciplined AI coding ships world-class production software, not the prompt-and-pray meme.

Two co-founders. One Chief Agentic Engineer at the keyboard. 1.7M+ files converted by real users in the browser. 18% Google search click-through rate. Zero paid acquisition. Robust, privacy-proof, production-grade evidence that disciplined AI coding ships world-class software in 2026.

Karpathy's definition (February 2026)

"Agentic, because the new default is that you are not writing the code directly 99% of the time, you are orchestrating agents. Engineering, because there is an art and science and expertise to it."

Andrej Karpathy, February 2026.

Vibe Coding is the meme. Agentic Engineering is the discipline.

Vibe Coding has been flattened into shorthand for shortcut, prompt-and-pray, speed over safety. DeckEdit sits on the other side of that line by construction, not by claim. Agentic Engineering is the disciplined practice of orchestrating AI coding agents under expert human direction: human-owned architecture, security review, verification, and taste, with agents doing 99% of the typing. Every feature is human-architected, every change human-reviewed before users see it. Nearly every coder in 2026 uses AI. Reaching DeckEdit's level of polish, privacy, and reliability requires this discipline. We are the existence proof.

Living proof: 10+ products shipped solo in 6 months

DeckEdit was co-founded by Keith Li and Roy Law, CEO and co-founder of TeamNote, a Y Combinator-backed company. Keith, the Chief Agentic Engineer, built the entire app using AI agents under his direct supervision and has shipped 10+ production-grade commercial products solo in 6 months using the same methodology, including DeckEdit, TranslationAI (a full SaaS with Stripe, multi-model orchestration, and an admin dashboard, shipped in 14 days), DesignThinkerLabs, ClassRail, AppToPub, DopaSprint, and Famiroot, plus the open source Auto-Academic-Paper on GitHub. This is not theory. It is a track record. It is the answer to anyone still claiming AI cannot code.

The proof is in the public record

Every number below is independently verifiable in Google Search Console, public analytics, or by inspecting the live site.

  • 18%

    Click-through rate recorded by Google Search Console. The web average is 3 to 5 percent. People do not just see DeckEdit, they click it.

    GSC, rolling 6 months ending May 2026

  • Position 5.0

    Average ranking across every indexed query. Top of page one, sustained, on a static SEO budget.

    GSC average, May 2026

  • 1.7M+

    Files converted by users directly in the browser. Zero of them ever touched our servers. We literally cannot see the content. Privacy-proof by architecture, not by promise.

    Cumulative through May 2026

  • $0

    Spent on paid acquisition. No Google Ads, no influencer deals, no growth team. Pure organic.

    Cumulative since launch

    1

    Chief Agentic Engineer at the keyboard. The other co-founder runs the business. AI agents do the implementation under direct human supervision, the textbook Karpathy definition of Agentic Engineering. A one-person army shipping world-class software. You do not need a team to ship at this level, you need the discipline. This is the rebuttal to two outdated mindsets at once, that AI cannot code, and that AI can only vibe-code.

    Engineering headcount, May 2026

Snapshots refreshed each release. Latest GSC export available on request.

The methodology

Agents do the implementation under direct human supervision. The Chief Agentic Engineer owns the architecture, security posture, product decisions, and the final call on what ships. Every change is reviewed by a human before it reaches users. No silent fallbacks, no fake features, no shortcuts. Not vibes. Not autopilot. Discipline plus delegation. This is what Karpathy means by the engineering half of the term, and it is why DeckEdit is robust, privacy-proof, and production-grade.

Frequently asked about Agentic Engineering

What is Agentic Engineering?

Agentic Engineering is the discipline Andrej Karpathy named in February 2026 for building production software by orchestrating AI coding agents under expert human direction. The agents do most of the typing. Humans own the architecture, the security review, the taste, and the call on what ships. It is the engineering side of AI coding, the opposite of the prompt-and-pray meme.

Is DeckEdit a Vibe Coding project?

No. DeckEdit is the opposite of Vibe Coding by construction. It is an Agentic Engineering project, and that distinction matters. Vibe Coding has come to mean prototyping by feel: ship fast, hope it works, accept the bugs. Every feature in DeckEdit is human-architected, every change human-reviewed, every release verified across all four SEO layers and the OCR, editor, and export pipelines before it reaches users. AI agents handle 99% of the typing under direct supervision. The result: 1.7M+ files converted, 100% local processing, zero security incidents.

Can two co-founders really ship production software with AI agents?

Yes, and DeckEdit is the proof. Two co-founders: Roy Law, CEO and co-founder of TeamNote, a Y Combinator-backed company, on the business side, and Keith Li as Chief Agentic Engineer at the keyboard. AI agents write the implementation under direct human supervision. 10+ shipped commercial products in 6 months. 1.7M+ files converted by real users in the browser. Zero paid advertising. The methodology is repeatable. The discipline is the part that takes practice.

What does Agentic Engineering mean for the people who use the product?

Faster iteration, smaller teams, lower prices, and higher quality than a conventional team of ten could ship in the same time. DeckEdit is free, has no account, runs entirely in your browser, and ships improvements every week. That cadence, at this level of robustness and privacy, is only possible because agents handle the volume and the founders handle the judgment.

Two founders. Agents wrote the code. Humans owned the architecture.

No growth team. No funding round. No marketing budget. Two co-founders practicing Agentic Engineering: Roy Law, CEO and co-founder of TeamNote, a Y Combinator-backed company, on the business side, and a Chief Agentic Engineer at the keyboard. A small, disciplined system that shipped world-class, privacy-proof software, and that real users found and kept using. This is what AI coding looks like when an engineer is at the keyboard, not just a prompter.

Open DeckEdit