Skip to content
VYNN AI · Claude · Agentic AI · Financial Modeling · EvaluationCite

VYNN AI Against Claude: Two Races on NVIDIA

Same words, same day. VYNN built a 10-tab model in 38 seconds and a 38-page report in 2 min 44 s. What was inside each, VYNN today, and how it got here.

A general agent can now build a financial model and write a research report. I wanted to see what a specialist still adds. On 15 October 2026 I gave VYNN and Claude the same two requests about NVIDIA, in the same words, and recorded both runs from start to finish.

Round one: “build me a financial model for nvda”

VYNN returned a 10-tab Excel model in 38 seconds, with 626 formulas. Claude, running Opus 5.5 on High in claude.ai, returned a 7-tab model in 7 min 24 s, with 418 formulas and 153 numbers typed into cells.
The difference that matters is inside the files. Both start from the same revenue, $130.5 billion. In Claude’s model it is a typed number with a 10-K citation beside it. In VYNN’s it is a formula that reads the figure from 784 rows of reported data stored in the workbook. VYNN’s fair value of $360.14 is a formula too: the average of its two DCF methods, $339.83 and $380.46.

Round two: “write me a research report on nvda”

VYNN finished a 38-page report, and the model behind it, in 2 min 44 s. Claude wrote a clear report of about 3,000 words in 5 min 16 s, without stopping to ask anything.
Both open with a bottom line. Then they part ways.
ClaudeVYNN
How it opensExplains the business from scratchA rating, a fair value and the evidence
ValuationThe P/E multipleTwo DCF models on its own spreadsheet, with a sensitivity grid
What the price assumesThree scenarios with round numbersA reverse DCF: about 17% revenue growth a year for ten years
Sources17 links13 dated pieces of evidence, and a valuation that traces back to the model
For a reader meeting NVIDIA for the first time, Claude’s report is the easier read. For someone deciding what the stock is worth, VYNN’s valuation can be followed back to the cell it came from.

How the races were run

Both were real production runs, one each, on the same day, with the same words. In round one Claude asked one clarifying question and was given the option it had marked Recommended. It also stopped once to ask permission to run a command. That wait is left out of the clocks on screen, so Claude’s own timer reads longer, about 8 min 36 s. The footage is sped up; the clocks on screen show real elapsed time. One run each is a demonstration, not a benchmark.

Why the specialist is faster

Claude has to decide everything on every request: what a model should contain, where each number comes from, how to lay it out. VYNN made those decisions once, in code. Its agent chooses the tools a question needs, but it cannot change how a valuation is computed. That is where the 38 seconds come from, and why each figure can be traced to its source.

VYNN today

VYNN is not only for stocks. One agent with 20 tools answers questions like these, all shown above:
  • Why is Bitcoin down about 27% while stocks sit near highs?
  • What do prediction markets expect from the Fed, and do Treasury yields agree?
  • How do the risks of TSLA, NVDA and the S&P 500 compare over three years?
  • Why did NVIDIA fall this week?
Everything you follow lives in one workspace: your portfolio, a page for each company with its financials, a thesis history that keeps every run, the full report, funds such as Vanguard’s S&P 500, Bitcoin, and the prediction markets you hold.

How it got here

VYNN started in July 2025 as “Value Your Next News”. The name sounds like “win”, which was the point. Then I rebuilt it, again and again:
  • Aug 2025: the first app.
  • Oct 2025: an invite-only beta.
  • Nov 2025: news, models in Excel and portfolios.
  • Apr 2026: its own website.
  • Aug 2026: the Living Market.
  • Sep 2026: the redesign.
  • Oct 2026: vynnai.com and the app as they are today.
You can ask VYNN a question at app.vynnai.com and read its dated research at vynnai.com. How it is built is in VYNN AI, Six Months Later and on VYNN’s page.

Cite this essay

If you build on or quote this essay, please cite it.

Zanwen Fu. 2026. “VYNN AI Against Claude: Two Races on NVIDIA.” zanwenfu.com, October 16, 2026. https://zanwenfu.com/blog/vynn_vs_claude

BibTeX
@misc{fu2026race,
  author       = {Fu, Zanwen},
  title        = {{VYNN} {AI} Against Claude: Two Races on {NVIDIA}},
  year         = {2026},
  month        = oct,
  howpublished = {\url{https://zanwenfu.com/blog/vynn_vs_claude}},
  note         = {Published October 16, 2026}
}

Published October 16, 2026.