🔍 Read the full analysis: Turning A Vibe-Coded PDF Editor Into Something Dependable on ThorstenMeyerAI.com
Get the latest gadgets 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 he built the first version of Docestra with AI coding agents in one day, then spent six days testing and revising it against real PDFs. His account describes defects involving saved annotations, encrypted files and redaction, and says the editor grew from 17 to 745 unit tests by version 1.5.1. Docestra is private, and its claimed test results and performance have not been independently verified.
Thorsten Meyer says he turned an AI-generated prototype of his PDF editor, Docestra, into a more extensively tested tool through six days of revisions, after testing real files exposed problems with saving, encrypted documents and redaction. In an account published November 13, he said development ran from the first working version on November 6 to version 1.5.1 on November 12; the private app is not available to download.
Meyer said an AI coding agent produced Docestra’s first working version on November 6, after he gave it a one-sentence request for an Acrobat-like tool. The local web app could open and search PDFs, add annotations and signatures, fill forms, rearrange pages and save files. He said the code was written using Anthropic’s Claude Code with Claude Opus, while OpenAI’s Codex reviewed it from the start. The app runs in a browser on his Mac, and, according to Meyer, files stay on the machine.
Testing a collection of 930 real PDFs revealed a saving problem: on a file generated by a common PDF tool, the saved document’s cross-reference table could be broken and markups could disappear. Meyer said he changed the save process to avoid compressed object streams. The same testing found that 52 of 69 encrypted PDFs did not permit annotations under restrictions set by their authors, while the PDF library could not re-encrypt edited files. Meyer said Docestra was changed to ask for confirmation, identify the restrictions and save an unprotected copy without overwriting the original.
Redaction required a different approach. The initial “white-out” feature only covered text; it did not remove the underlying content. Meyer said an independent AI review identified 14 problems in the first review round and 13 in the second for an early feature designed to remove content in place. He replaced it with a method that converts affected pages to images, applies redactions to their pixels and adds a searchable text layer that omits text intersecting the marked areas. He reports that testing on 917 PDFs found zero leaks. These are results from Meyer’s own testing, not an independently published evaluation.
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
Why Real PDFs Changed the Build
The account illustrates a gap between a software demo and a tool people can trust with everyday documents. A prototype can appear functional on a few sample files while still losing annotations, mishandling encryption or giving users the false impression that covered text has been removed. Those failures matter when PDFs contain contracts, account numbers or other sensitive information.
Meyer’s experience also shows that generating code quickly does not eliminate the work of testing decisions and their consequences. His redaction revision traded the original approach for a simpler image-based design, while his encryption change avoided silently replacing protected files with unprotected ones. He says the editor’s automated unit tests increased from 17 to 745 by version 1.5.1, with tests added for problems found during review. That growth is a measure of the project’s testing work, not evidence on its own that the editor is secure or suitable for every use.
The account does not establish that AI-built software generally requires a particular amount of repair. It documents one developer’s process and test results. Meyer also says he has not measured time savings or cost savings, so the project offers no quantified comparison of AI-assisted development with another approach.
Products Worth Considering
LG UltraGear 27G810A-B 27-Inch 4K UHD IPS Gaming Monitor
Kindle Paperwhite Signature Edition 32GB E-Reader, 7-Inch Display, Seaglass Green
Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite
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 a desired result in plain language to an AI model and letting it write the code. He says he began Docestra with a one-sentence brief on November 6, 2026; the first version was committed that evening at 21:47 Central European Time. Its early features included viewing, annotation, page management, form filling and saving PDFs.
The work described in his account covers releases from version 1 to version 1.5.1, with the latter completed November 12. During that period, file testing, code review and decisions about library licences led to changes beyond the original brief. Meyer says he excluded fonts shipped under the GPL-2.0 licence because the project’s rules do not permit GPL code, and later checked licences before adding libraries for text recognition and editing.
Docestra remains a personal, private tool rather than a publicly downloadable product. Meyer said development on version 2.0 began November 13 and was not finished when he published his account.
“Vibe coding gets you a working tool quickly. Turning it into a tool you can rely on is the real work.”
— Thorsten Meyer, describing the project’s development
What the Tests Do Not Establish
The reported test results come from Meyer’s own account. The article does not provide an independently conducted audit, detailed test methodology or access to Docestra, which he says is private and unavailable to download. It is not clear how the 930-file collection was assembled, what range of PDF producers and document types it covers, or how the redaction leak tests were conducted.
The claim of zero leaks on 917 PDFs describes the results of that test set, not a guarantee that all redactions will be safe in other files or circumstances. The account also does not establish whether the app has undergone a security assessment, how it performs for other users, or whether its local-file handling has been independently checked. Meyer says he has not measured financial or time savings, and version 2.0 remains unfinished.
Version 2.0 Remains in Development
Meyer said work on Docestra version 2.0 began November 13, 2026, the day his account was published, and that it was not complete. He did not provide a release date or say whether he plans to make the software public. Until further details are available, the next reported milestone is the completion of that development work; the account does not specify what changes version 2.0 will include.
Key Questions
What is Docestra?
Docestra is a local web-based PDF editor that Meyer says he built with help from AI coding agents. Its features include annotations, signatures, form filling, page rearrangement and saving PDFs. Meyer says files remain on his Mac, and the tool is not available to download.
What problems appeared during testing?
Meyer reports that a save process could make annotations disappear from some files, that many of the encrypted PDFs he tested did not permit annotation, and that the original white-out tool covered text without removing it. He says he revised the save and encryption workflows and replaced the redaction design.
Does the redaction test prove Docestra is safe for sensitive documents?
No. Meyer reports zero leaks in tests on 917 PDFs, but the results have not been independently verified, and the account does not provide enough methodology to establish performance on all documents. The result should not be treated as a general guarantee.
Can the public download Docestra or version 2.0?
No public download is described. Meyer says Docestra is private and that version 2.0 was still in development as of November 13, 2026; he did not announce a release date.
Source: ThorstenMeyerAI.com
