I Built A PDF Editor With AI, And Reliability Made The Difference
KIDieser Beitrag wurde mit Unterstützung künstlicher Intelligenz (KI) erstellt.

🔍 Read the full analysis: I Built A PDF Editor With AI, And Reliability Made The Difference on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Thorsten Meyer says an AI coding agent produced the first working version of his PDF editor, Docestra, on Nov. 6, 2026. Subsequent testing across real files exposed save, encryption and redaction issues; by version 1.5.1, the project had 745 unit tests, but it remains a private tool and its time or cost savings have not been measured.

Thorsten Meyer says AI coding agents built the first working version of his PDF editor, Docestra, in a day, but testing real documents over the following six days exposed problems with saved markups, encrypted files and redaction. In an account published Nov. 13, Meyer says the private, local web app reached version 1.5.1 on Nov. 12, after revisions and tests; it is not available to download.

Meyer says he gave an AI coding agent a one-sentence brief on Nov. 6, 2026, to build a PDF editor he could use. The first working version was committed that evening. It could open and search PDFs, add annotations and signatures, fill forms, rearrange pages and save files. The code was written using Anthropic’s Claude Code with Claude Opus, while OpenAI’s Codex reviewed it, according to Meyer.

The first major reliability check covered 930 real PDFs. Meyer says it found that saving some files could produce a broken cross-reference table, causing markups to disappear. He changed the save process to avoid compressed object streams. Testing also found that 52 of 69 encrypted PDFs did not permit annotations under restrictions set by their authors. Because the PDF library could not re-encrypt edited files, Docestra was changed to request confirmation before unlocking and to save an unprotected copy without overwriting the original.

Redaction required a separate redesign. An early white-out tool only covered text visually, leaving the underlying content in the PDF. Meyer says an independent AI review identified 14 problems in the first review round and 13 in the second for an initial real-redaction design. He then switched to converting pages with redactions into images, covering marked areas in the pixels and adding a searchable text layer that excludes lines touching a redaction. Meyer reports that a test across 917 PDFs found no leaks. He also says the software had 745 automated unit tests in version 1.5.1, up from 17 in its first version.

At a glance
reportWhen: Published November 13, 2026; developmen…
The developmentMeyer published an account of six days of testing and revisions to Docestra, describing reliability problems found after an AI coding agent built its first working version.

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

  1. Nov 6One-sentence brief; first working version committed the same evening
  2. Nov 6Test over 930 real PDFs finds a save bug that dropped markups
  3. Nov 752 of 69 encrypted PDFs forbid annotations: unlock only into a new copy
  4. Nov 8Real redaction: review finds 14, then 13 problems; simpler design, 0 leaks in 917 PDFs
  5. Nov 12Version 1.5.1: 745 unit tests, 32 browser test suites
  6. Nov 13Version 2.0 starts: one shared PDF engine for our AI agents (in development)
Working is not the same as reliable. The first version worked on the files I tried. Real files decided what had to change.

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-File Testing Changed Docestra

Meyer’s account illustrates a distinction between a tool that performs its advertised tasks on a few sample documents and one that handles varied files without losing edits or exposing information. A broken save or a visual-only redaction can have consequences beyond inconvenience, particularly when PDFs contain contracts, account details or other sensitive material. His reported tests describe specific risks that a simple demonstration might not reveal.

The account is also a case study in the limits of AI-assisted software development. The agent produced a usable starting point quickly, but Meyer says testing, review, licensing decisions and rework took longer than building some of the features. That is his experience, not a measured comparison of development methods. He says he has not measured time or cost savings, so the project does not establish that AI coding made the overall work faster or cheaper.

For readers choosing software, the details matter as much as the feature list: whether edits persist, how encrypted files are handled, and whether redaction removes underlying content. Docestra runs locally on Meyer’s Mac, and he says files stay on that machine. Its private status, however, means readers cannot independently download and test the application from the account alone.

Products Worth Considering

Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite

Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite

Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C - Graphite

Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C – Graphite

As an affiliate, we earn on qualifying purchases.

From One-Day Build to Version 1.5.1

Meyer defines “vibe coding” as building software by describing desired features in plain language to an AI model, which then writes code. Docestra began as a TypeScript local web app on Nov. 6. Its first version included viewing, search, annotation, page management, form filling and saving. Meyer’s account follows changes through version 1.5.1, committed on Nov. 12.

Several decisions went beyond adding features. Meyer says the editor does not bundle fonts covered by the GPL-2.0 licence because his project rules exclude GPL code. He also says he reviewed licences for later libraries added for text recognition and editing. The account describes a private development project rather than a public product launch: Docestra is not available to download.

Meyer says version 2.0 development began on Nov. 13 and was not finished at the time of publication. The figures and test outcomes in the account are his reports about his own software and test files; they are not an independent assessment of PDF editors generally.

“Vibe coding gets you a working tool quickly. Turning it into a tool you can rely on is the real work.”

— Thorsten Meyer

What the Test Results Do Not Establish

The reported test results do not show how Docestra performs across every PDF or operating environment. Meyer says the redaction test found zero leaks across 917 PDFs, but the account does not provide an independent audit, the full test protocol or enough detail to assess how closely those files represent other users’ documents. The reported result should not be read as a guarantee that redaction is risk-free.

Meyer has not published the editor for others to inspect or test, and he says time and cost savings have not been measured. The account also does not specify when version 2.0 will be complete or what features it will include. It remains unclear whether Docestra will ever be released publicly.

Version 2.0 Remains in Development

Meyer says work on version 2.0 began Nov. 13, 2026, the day his account was published, but the release was unfinished. He has not provided a completion date or a public release plan. For now, the stated next step is continued development of a tool he uses privately.

Any wider assessment will depend on further details, including whether Docestra becomes available for independent testing and whether Meyer publishes additional results on its reliability, security and performance. Until then, the reported figures describe his own test runs and development process.

Key Questions

What is Docestra?

Docestra is a PDF editor that Thorsten Meyer says he built with AI coding agents. It runs as a local web app on his Mac and, according to Meyer, keeps files on the machine.

When was the first working version built?

Meyer says the first working version was committed on Nov. 6, 2026, at 9:47 p.m. Central European Time, the same day he gave the AI coding agent its brief.

What problems did testing find?

Meyer reports that tests across 930 PDFs found a save issue that could make markups disappear. He also describes limits involving encrypted PDFs and an early white-out tool that covered text without removing it from the file.

Is Docestra available to download?

No. Meyer says Docestra is private and not available to download. His account does not announce a public release.

Has Meyer shown that the AI-built editor saves time or money?

No. Meyer explicitly says he has not measured time or cost savings. He describes his own development experience rather than a measured comparison with other approaches.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Vocal-strain load tracking for working singers

A new app prototype monitors vocal strain for professional singers, aiming to prevent injury by providing early alerts after each performance.

ShinyHunters · The New APT Model.

Analysis of ShinyHunters’ evolving operational model, AI-enabled tactics, and implications for enterprise security since 2020.

Best 4-bay NAS for Teams: How to Avoid Paying Premium Prices for Old Hardware

Discover the best 4 bay NAS for teams in 2026. Find top picks for performance, value, and ease of use tailored to your team size and needs.

The High-End PC And Workstation Tax

Memory costs have surged in 2026, transforming PC building economics. DIY builders face higher prices, while OEMs leverage bulk buying. Details inside.