🔍 Read the full analysis: The Real Project Started After I Vibe-Coded My PDF Editor on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Thorsten Meyer says an AI coding agent built the first working version of his private PDF editor, Docestra, on Nov. 6, 2026. Tests on real documents then exposed saving, encryption and redaction problems; by version 1.5.1, released Nov. 12, the project had grown from 17 to 745 automated unit tests. Meyer has not measured time or cost savings, and the app is not publicly available.
Thorsten Meyer says an AI coding agent produced a working first version of his PDF editor, Docestra, in one day, but tests on real documents uncovered problems that led to six days of revisions. In an account published Nov. 13, he describes how the private app reached version 1.5.1 on Nov. 12, adding fixes and tests after issues surfaced in PDF saving, encrypted files and redaction.
Meyer says he gave the coding agent a one-sentence brief on Nov. 6: build an Acrobat-style editor he could use. The first version was committed at 21:47 Central European Time. It could open and search PDFs, add highlights and annotations, fill forms, sign, rearrange pages and save files. The app is a local web application written in TypeScript and runs in a browser on Meyer’s Mac; he says documents stay on his machine.
A test run involving 930 real PDFs found a saving bug: a common PDF format could produce a broken cross-reference table, causing markups to disappear from the saved file. Meyer says he changed the save process to avoid compressed object streams. Testing also found that 52 of 69 encrypted PDFs did not permit annotations under their authors’ restrictions. Because the PDF library could not re-encrypt edited files, Docestra was changed to ask for confirmation, identify the restrictions and save an unprotected copy without overwriting the original.
Redaction required a larger redesign. The initial white-out feature only covered text visually, leaving the underlying content in the PDF. Meyer says an independent AI review identified 14 problems in the first version of a true redaction feature and 13 in a second review. He then switched to rendering redacted pages as images, applying the redactions to the pixels and adding a searchable text layer that excludes text touching marked areas. Meyer reports that a test on 917 PDFs found no leaks. He also says the project’s automated unit tests increased from 17 in the initial version to 745 in version 1.5.1.
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
Prototype Gains Tests and Safeguards
The account illustrates the difference between producing a demonstration and building software for repeated use. A tool may appear to work on a handful of documents while failing on files with different structures, permissions or sensitive information. In Meyer’s tests, the problems were not cosmetic: one could remove edits from a saved file, while the early white-out tool could leave supposedly hidden text intact.
For readers using AI-generated tools, the practical point is that reliability depends on testing against representative files and making deliberate choices when a quick fix would create a new risk. Meyer’s approach included preserving original encrypted documents, documenting restrictions, using an outside model for review and adding automated tests for discovered failures. These are details from one personal project, not evidence that the same process or results apply to other software.
Meyer says he has not measured time or cost savings. The project therefore offers a development account, not a quantified claim that AI coding reduced the resources needed to build or maintain the editor.
Popular right now
Vitamix 5200 Blender with 64 oz Container, Black
KitchenAid Classic Series 4.5-Quart Tilt-Head Stand Mixer, Onyx Black
eufy C28 Robot Vacuum and Mop for Pet Hair, 15,000 Pa
As an affiliate, we earn on qualifying purchases.
From One-Day Build to Version 1.5.1
Meyer defines vibe coding as building software mainly by describing desired features to an AI model in ordinary language and letting it write the code. He says Docestra’s code was produced with Anthropic’s Claude Code, running Claude Opus, while OpenAI’s Codex reviewed the project from the start. The app is private and, according to Meyer, is not available to download.
The first version included a broad set of editing functions, including annotations, page rotation, combining and extracting pages, printing and undo. Meyer’s account then follows revisions from Nov. 6 through version 1.5.1 on Nov. 12. He says version 2.0 has been in development since Nov. 13 and is unfinished. The source also describes a licensing decision: Meyer says he excluded fonts distributed under GPL-2.0 because the project’s rules did not allow GPL code, and later checked the licences of libraries considered for text recognition and editing.
“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
Limits of the Reported Tests
The test results are reported by Meyer, and the source does not provide the test files, detailed methods or an independent verification of the findings. The reported result of zero leaks in 917 PDFs applies to the redesigned redaction approach and that test set; it does not establish that every document or redaction scenario is safe.
It is also unclear how Docestra performs across operating systems or with PDFs outside Meyer’s test collection. The app is private, so readers cannot independently try it or inspect its code based on the source account. No time or cost savings have been measured, and version 2.0 remains unfinished.
Version 2.0 Remains in Development
Meyer says work on Docestra version 2.0 began on Nov. 13, 2026, but he does not give a release date or describe a finished feature list. The next reported milestone is therefore the completion of that development work; its scope and timing remain unspecified.
For now, the account ends with version 1.5.1, a private tool Meyer says he uses in his own workflow. He has not announced a public release, and the source does not say whether he plans to make Docestra available to others.
Key Questions
What is Docestra?
Docestra is a private PDF editor Thorsten Meyer says he built with help from AI coding agents. It runs as a local web app on his Mac and is not available to download, according to his account.
What problems did testing uncover?
Meyer reports that tests on real PDFs found a save bug that could make markups disappear, restrictions on annotating some encrypted files, and a white-out feature that covered text without removing it from the document.
How did Meyer change the redaction feature?
After two rounds of independent AI review identified problems, Meyer says he changed the design. Redacted pages are rendered as images with marked areas removed from the pixels, alongside a searchable text layer that excludes text touching those areas.
Has Docestra been independently verified or released?
The source account does not provide independent verification of Meyer’s test results, and he says the app is private and unavailable for download. Version 2.0 is still in development, with no release date given.
Did the project save time or money?
Meyer says he has not measured time or cost savings. His account describes the build and testing process but does not quantify financial or productivity gains.
Source: ThorstenMeyerAI.com
