🔍 Read the full analysis: Was Building A PDF Editor With AI The Easy Part? on ThorstenMeyerAI.com
Get hardware and tech essentials 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 a working first version of his private PDF editor, Docestra, on Nov. 6, 2026. Over the following six days, tests across real PDFs and independent code reviews uncovered problems that led to changes in saving, encrypted-file handling and redaction; the tool is not publicly available, and its time and cost savings have not been measured.
Thorsten Meyer says an AI coding agent built the first working version of Docestra, his private PDF editor, on November 6, 2026, but testing and reworking it over the next six days exposed problems that a quick demonstration had not revealed. In an account published November 13, Meyer said tests involving hundreds of real files and reviews of its redaction feature led to changes in how the tool saves and handles PDFs—showing that a fast first build did not establish that the software was reliable.
Meyer describes Docestra as a local web app written in TypeScript and used in a browser on his Mac. He says files stay on that machine. The initial version supported viewing and searching PDFs, adding annotations and signatures, filling forms, changing page order, combining or extracting pages, and saving. Meyer says the first working version was committed at 21:47 Central European Time on November 6. By November 12, the project had reached version 1.5.1.
A robustness run over 930 real PDFs found a save problem: in some files from a common PDF generator, the PDF library could produce a broken cross-reference table, causing markups to disappear after saving. Meyer says he changed the save process to avoid compressed object streams. His tests also included 69 encrypted PDFs. According to his account, 52 did not permit annotations under restrictions set by their authors, while the library could not re-encrypt edited files. Docestra was changed to ask for confirmation before unlocking those files and to save an unprotected copy without overwriting the original. Meyer reports that all 69 test files then unlocked.
Redaction required a more substantial design change. The initial “white-out” tool covered content without removing the underlying text. Meyer says an independent AI review found 14 problems in the first review round of the first real-redaction design and 13 in the second. He abandoned that design. The replacement turns pages with redactions into images, removes the marked text from the image, and adds a searchable text layer that omits text intersecting the marks. Meyer reports zero leaks in a test of 917 PDFs; that result describes his test set, not a guarantee that all documents or cases are safe.
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
The account illustrates a practical distinction for people using AI-generated software: a tool can perform its advertised tasks on a few sample files while still failing on documents encountered in ordinary work. Here, a defect in saving could make annotations disappear, and a tool that only covered sensitive text could leave that text in the file. These are not cosmetic issues for a PDF editor used with contracts or other records.
Meyer says the project’s automated unit tests grew from 17 in the first version to 745 in version 1.5.1, with tests added for problems found during review and testing. That gives readers a concrete account of how the project changed, but it is not an independent assessment of the editor’s overall safety or quality. Meyer also says he has not measured time or cost savings, so the account does not establish that AI development made the completed tool cheaper or faster to maintain.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Clicky Switches
Apple MacBook Pro 16.3-inch with PROLED XDR Display, Silver
Apple MacBook Pro 14.2-Inch Laptop with M6 Chip, Space Black
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 plain language and having the model write code. He says Anthropic’s Claude Code, running Claude Opus, wrote Docestra’s code, while OpenAI’s Codex reviewed it from the start. The initial brief was to build a personal Acrobat-like editor. Its first version included a broad set of features, but the account’s focus is what surfaced after that first build.
In Meyer’s timeline, development ran from version 1 on November 6 to version 1.5.1 on November 12. He says he chose to keep improving Docestra because it already fit his workflow and because the tests and reviews could accumulate as safeguards against repeat defects. He also describes checking licences before including additional libraries, after discovering that a PDF viewer library shipped fonts under GPL-2.0, a licence his project rules excluded.
The account is Meyer’s own report on ThorstenMeyerAI.com, not an independent product evaluation. Docestra is described as private and unavailable to download. Meyer says work on version 2.0 began November 13 and is not finished.
“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 counts and results come from Meyer’s account; the source provides no independent replication, detailed test methodology, or public access to Docestra. The reported zero leaks across 917 PDFs does not show how the replacement redaction method performs on documents outside that test set or under other conditions. The account also does not specify the full range of PDF generators, encryption settings, or document features represented in its test files.
Docestra remains private, so readers cannot independently inspect or test the software. Meyer says he has not measured development time or costs saved, and the source gives no figures for either. It also does not establish whether the product will be released or how version 2.0 will differ from the current version.
Version 2.0 Remains in Development
Meyer says development of Docestra 2.0 began November 13, 2026, the date of his report, and that the version is unfinished. He has not announced a release date or said that the editor will become publicly available. For now, the next stated milestone is continued development; what features or validation work will be included in version 2.0 remains unspecified.
The account’s practical next step is continued testing and review as the tool changes. Whether Docestra’s fixes hold up beyond Meyer’s reported test files—and whether the project will publish enough detail for others to verify them—remains open.
Key Questions
What is Docestra?
Docestra is a private PDF editor that Thorsten Meyer says he built with AI coding tools. It runs locally in a browser on his Mac, and he says files stay on that machine. It is not available to download.
How quickly was the first version built?
Meyer says the first working version was committed at 21:47 Central European Time on November 6, 2026, the day he gave the coding agent a one-sentence brief. He describes it as functional, but says later tests revealed issues that needed further work.
What problems did testing reveal?
Meyer reports that a save defect could cause markups to disappear, that some encrypted PDFs did not allow annotations under their authors’ restrictions, and that an initial white-out tool left the underlying text in place. He says the editor was revised to address those issues.
Does the redaction test prove Docestra is safe for every PDF?
No. Meyer reports zero leaks in a test of 917 PDFs, but that result applies to the files tested. The source does not provide independent verification or establish how the method performs across all PDF types and conditions.
Has Meyer measured time or cost savings?
No. Meyer explicitly says he has not measured time or cost savings. His report describes the rapid first build and the subsequent testing and rework, but does not quantify the overall savings.
Source: ThorstenMeyerAI.com
