🔍 Read the full analysis: A More Reliable PDF Editor Took More Than Vibe-Coding on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Thorsten Meyer says an AI coding agent produced the first working version of his private PDF editor, Docestra, on Nov. 6, 2026. Over the following six days, testing hundreds of real files and reviewing its redaction design exposed reliability and security problems; Meyer says the work expanded the tool from 17 to 745 automated unit tests. He has not measured time or cost savings, and Docestra is not available to download.
Thorsten Meyer says an AI coding agent built the first working version of his PDF editor, Docestra, in one day, but testing real documents over the next six days exposed problems with saving, encrypted files and redaction. In an account published Nov. 13, Meyer said revisions through version 1.5.1 expanded the app’s automated unit tests from 17 to 745; he has not measured whether the project saved time or money.
Meyer said he gave an AI coding agent a one-sentence brief on Nov. 6, 2026, to build a personal Acrobat alternative. The first working version was committed that evening at 21:47 Central European Time. It could open PDFs, add highlights, drawings, notes and signatures, fill forms, change page order and save files. The local TypeScript web app runs in a browser on his Mac, and Meyer said documents stay on the machine.
The more consequential findings came from testing beyond the small number of files used during initial development. A run over 930 real PDFs exposed a save problem: for a file produced by a common PDF generator, the PDF library could write a broken cross-reference table, causing markups to disappear from the saved document. Meyer said he addressed this by saving without compressed object streams.
Encrypted files created a separate decision. Meyer reported that 52 of 69 encrypted PDFs in his test did not permit annotations under restrictions set by their authors, and that the library could not re-encrypt a file after editing. He said Docestra was changed to require confirmation before unlocking such a file, identify the restrictions and save an unprotected copy without overwriting the original. He reported that all 69 test files then unlocked.
Docestra · Nov 6–13, 2026
From one prompt to a tool I can rely on
What the first version gave me, what real files broke, and what it took to fix it, from my project log.
What the tool rests on (schematic)
Built bottom up. The top layer is a plan, not a finished feature.
What happened, by date
- Nov 6One-sentence brief; first working version committed the same evening
- Nov 6Test over 930 real PDFs finds a save bug that dropped markups
- Nov 752 of 69 encrypted PDFs forbid annotations: unlock only into a new copy
- Nov 8Real redaction: review finds 14, then 13 problems; simpler design, 0 leaks in 917 PDFs
- Nov 12Version 1.5.1: 745 unit tests, 32 browser test suites
- Nov 13Version 2.0 starts: one shared PDF engine for our AI agents (in development)
Source: Docestra project records (development log, plans for versions 1 and 2.0, commit history), Nov 6–13, 2026. The layer stack is a schematic, not a measurement.
Graphic: Powered by Thorsten Meyer AI
Testing Changed the Editor’s Design
The account shows the gap between a prototype that performs advertised actions and a tool that handles varied documents without silently losing work or misrepresenting what it has changed. In Meyer’s tests, a save defect was not apparent in the initial demonstration, while encrypted documents raised a question the original brief had not addressed: what should the app do when editing means removing document protection?
Redaction made the stakes clearer. Meyer said the initial “white-out” feature covered text visually but left the underlying content in the PDF. That distinction matters when a user believes a name or account number has been removed. After an independent AI review identified 14 problems in one round and 13 in a second for the first real-redaction design, Meyer abandoned that approach rather than continue patching it. The replacement rasterizes pages with redactions, applies marks to the image and adds a searchable text layer that omits lines touching a mark. Meyer reported no leaks in a test of 917 PDFs; that result describes his test set, not a guarantee for other files.
The work also illustrates that AI-generated code does not remove the need for testing, design choices or licence checks. Meyer said the PDF viewer library included fonts under the GPL-2.0 licence, which his project rules excluded, so he did not ship them. He said he later reviewed licences before adding libraries for text recognition and editing. Those are practical considerations for people building tools that handle contracts, personal records or other sensitive files.
Products Worth Considering
Dell Pro 14 Plus P1425 14-Inch Portable Monitor
Canon SELPHY CP1500 Wireless Dye-Sublimation Photo Printer Bundle with 108 Sheets of Paper and 3 Ink Cartridges
FLEXISPOT Electric Adjustable Workbench, 48 x 30 in, Maple
As an affiliate, we earn on qualifying purchases.
From One-Day Prototype to Version 1.5.1
Meyer defines “vibe coding” as building software mainly by describing desired features in ordinary language and having an AI model write the code. He said Claude Code, running Claude Opus, generated Docestra’s code, while OpenAI’s Codex reviewed it from the start. The first version included viewing, searching, annotation, form filling and page-management features, along with printing, undo and saving.
Between Nov. 6 and Nov. 12, Meyer released versions through 1.5.1, adding fixes and tests as issues emerged. He said each real problem found in review now has a test intended to catch a return of the same failure. The project’s unit-test count grew from 17 in the first version to 745 in version 1.5.1. Meyer’s account says the tests and rework took longer than building for several packages, but does not provide measured hours or a comparison with conventional development.
Meyer chose to keep improving Docestra because it fit his own workflow and because fixes and tests accumulated in the tool he used. He describes the project as personal software, not a public product: Docestra is private and unavailable to download.
“Vibe coding gets you a working tool quickly. Turning it into a tool you can rely on is the real work.”
— Thorsten Meyer, describing his conclusion from building Docestra
Limits of the Reported Tests
The test results are Meyer’s own reported findings; the source account does not provide independent verification, a full test protocol or enough detail to assess how representative the PDF samples were. The reported zero leaks in 917 files should not be read as proof that every redaction in every document is safe. It is also not clear how the app behaves across operating systems or configurations beyond Meyer’s Mac-based setup.
Meyer said he has not measured time or cost savings, so the project does not establish that this approach was faster or cheaper overall than other development methods. Docestra is private and unavailable to download, limiting outside users’ ability to inspect or reproduce the results. Version 2.0 was in development as of Nov. 13 and was not finished.
Version 2.0 Remains in Development
Meyer said development of Docestra 2.0 began Nov. 13, 2026, the day his account was published, and that the release was not complete. He did not give a launch date or say whether the software would become publicly available. The next reported milestone is therefore completion of that ongoing development, but its features and timing remain unspecified.
For now, the account documents a private project rather than a downloadable product or independently evaluated release. Meyer’s next steps, as described, are continued development and testing; no public release schedule or further test results were provided.
Key Questions
What is Docestra?
Docestra is a private PDF editor Meyer built for his own use with AI coding agents. He says it runs locally in a browser on his Mac and is not available to download.
What did testing reveal about saving PDFs?
Meyer said a run over 930 real PDFs found that some saved files could have a broken cross-reference table, making markups disappear. He said saving without compressed object streams addressed the issue.
Did Docestra remove the text beneath its original white-out marks?
No. Meyer said the original white-out feature only covered content visually, leaving the text in the PDF. He later replaced the first real-redaction design after reviews found problems.
Does the reported redaction test prove the feature is safe for every file?
No. Meyer reported zero leaks in a test of 917 PDFs, but the account does not establish that the test covered every file type or use case. It is a reported test result, not a universal guarantee.
Has Meyer shown that vibe coding saved time or money?
No. Meyer said he has not measured time or cost savings. He reported that testing and rework took longer than building for several packages, but supplied no overall comparison.
Source: ThorstenMeyerAI.com
