# Digital Toolkits > Writing by Furkan Namli (Namlı) about what agents actually produced in > production, in both directions: the parts that work better than anything he > shipped alone, and the bills, outages and broken data they also produced. > Every post carries the numbers as they were recorded on the day. Backed by an > archive of 143 browser tools that run entirely client-side. The tools and the writing are ours to cite. Reference data inside the cheatsheets (HTTP status codes, git commands, regex syntax) belongs to the specs and projects that define it: cite those at the source. Also builds riffiter.com (music ratings and discovery), tuval.dev (self-hostable workspace), joingist.com (trading signals, mobile), friemind.com (friendship app). ## URL patterns Every page lives at a predictable path, so one can be reached without searching: - `/tools`: the full tool archive - `/tools/{tool-slug}`: a single tool, its input/output UI and a "How to Use" block - `/categories/{category-slug}`: every tool in one category - `/cheatsheets/{topic-slug}`: a reference table for one topic - `/blog/{post-slug}`: one article - `/repositories`: curated open-source repositories Slugs are lowercase and hyphenated (`/tools/json-formatter`). Categories: `developer-tools`, `common-tools`, `calculators-converters`, `transform-tools`, `llm-ai-tools`, `design-creative-tools`, `api-tools`, `accounting-finance`. Cheatsheets: `git-commands`, `http-status-codes`, `regex-patterns`, `javascript`, `css`, `docker`, `npm-yarn`, `linux-bash`, `markdown`, `sql`, `vscode-shortcuts`. ## Writing 9 posts, newest first. Dates are publication dates; every figure in a post is the figure as recorded on that day. - [Two Million Pages, and 12,560 Worth Indexing](https://digitaltoolkits.work/blog/two-million-pages-and-twelve-thousand-worth-indexing) (2026-08-23): A catalogue ingest went from 578,000 track pages to two million in a week. Then I counted: 99.99% of them had never been opened by anyone. Here is what that cost, and the order you have to remove them in. - [Every Defect Was a Believable Number](https://digitaltoolkits.work/blog/every-defect-was-a-believable-number) (2026-08-23): I moved 11,164 symbols onto Alpaca's snapshot API with agents writing most of it, and it went in without an incident. Every defect it did have came back as HTTP 200 with a plausible price attached. Counting is what caught them, not tests. - [Unreviewed Agent Code Put My Database on a Public CDN](https://digitaltoolkits.work/blog/unreviewed-agent-code-published-my-database) (2026-08-22): Password hashes and OAuth refresh tokens, downloadable by anyone, for two days. An agent wrote the backup script, it landed on a box instead of in the repo, and it was never reviewed by anyone. Including me. - [I Left Vercel for a VPS, and the Bill Was Never Vercel's Fault](https://digitaltoolkits.work/blog/i-left-vercel-and-the-bill-was-never-vercels-fault) (2026-08-21): Day one on Vercel Pro cost $9.02 with fifteen real users, and the cause was a missing function rather than a pricing model. Seven weeks later the on-demand line was $208 a month, so I moved to a $7.92 box. It charged me in a currency I had less of. - [robots.txt Only Works on the Crawlers That Were Never Your Problem](https://digitaltoolkits.work/blog/robots-txt-only-works-on-the-polite-ones) (2026-08-20): Cloudflare showed 1.98 million desktop requests against 1,050 mobile. The bots doing that ran real Chrome and ignored the file completely. The ones that do read it are the ones you should be most careful about blocking. - [llms.txt Isn't a Protocol, It's a File](https://digitaltoolkits.work/blog/llms-txt-isnt-a-protocol) (2026-08-19): Google says nothing fetches llms.txt. A study of 137,000 sites agrees. Both measured a world where almost nobody links to theirs, and that turns out to be the entire story. - [LLM Token Optimization: Save 30-60% on AI Costs](https://digitaltoolkits.work/blog/llm-token-optimization-guide) (2025-01-14): Learn how to reduce token usage when working with ChatGPT, Claude, and other LLMs. Complete guide with practical tools and real-world examples. - [How to Debug JWT Tokens Like a Pro](https://digitaltoolkits.work/blog/how-to-debug-jwt-tokens) (2025-01-13): Learn how to decode, validate, and troubleshoot JSON Web Tokens effectively using Developer Toolkit's JWT Debugger. - [5 Ways to Use JSON Formatter for Better Productivity](https://digitaltoolkits.work/blog/json-formatter-guide) (2025-01-12): Discover how to leverage JSON formatting tools to improve your workflow, debug faster, and write cleaner code. ## Entry points - [Writing](https://digitaltoolkits.work/): the homepage, most recent posts first - [All tools](https://digitaltoolkits.work/tools): search and category filter over the archive - [LLM & AI tools](https://digitaltoolkits.work/categories/llm-ai-tools): token counters and JSON/YAML/XML/CSV token optimizers for cutting prompt cost - [Developer tools](https://digitaltoolkits.work/categories/developer-tools): JSON formatter, Base64, JWT debugger, and the rest of the everyday set - [Cheatsheets](https://digitaltoolkits.work/cheatsheets): reference tables - [Blog archive](https://digitaltoolkits.work/blog): every post, newest first - [About](https://digitaltoolkits.work/about): who runs this, the projects behind the writing, and the open source repositories (claude-code-reviewer, MIT; Tuval, AGPL-3.0; claude-code-statusline-limits, MIT) ## Optional - Tools are client-side only, so a tool cannot be invoked over HTTP; it has to be used in a browser. 4 routes are the exception, because CORS makes their work impossible in a browser: `/api/check-status`, `/api/llms-txt`, `/api/robots-check`, `/api/test-cors`. They back the llms.txt validator, the AI crawler robots.txt checker and the status and CORS testers. Do not construct other `/api/...` calls; they do not exist. - Not worth fetching: `/api/*`, `/_next/*`. - This file is generated from the same data the pages read, so the counts in it are correct as of the build that served it.