Aervio
Pre-launchA drone-services marketplace — it turns “inspect my roof panels” into an operator-ready technical brief. Not taking orders yet — the site runs a waitlist.
See aerv.io →Products & tools
What I build
Products first — live and taking users. Below them, small sharp tools from the night shift.
Paid.
The e-mail with the download links goes to the address you gave at checkout, usually within a minute. The links work for 30 days, and the invoice follows in a separate message.
If the e-mail has not arrived within an hour, check your spam folder and write to kontakt@solertio.pl.
Products
Lucevio
Public betaYour personal colour, makeup, hair and outfit report — generated from a few photos. A 12-type colour system, identity-preserving visuals, a magazine-style report.
Open the app →Solertio Listing
Live · BYOKA real-estate toolkit: location due-diligence built from official Polish geodata (ULDK, MPZP, flood & price registers), plus AI staging and listing copy for the sale.
Open the tool →Tools
Solertio Mirror
Turn any live website into a local, editable copy — assets, fonts, animations and all.
How it works
- 01
Discover
Detects the framework, routes, asset inventory and dependencies.
- 02
Capture
Headless Chromium records network traffic, scrolls to trigger lazy-loading, saves fonts, video,
.glband.hdr, and rewrites URLs to relative paths. - 03
Rebuild
A clean recreation: design tokens, section by section.
- 04
Serve
A local server with correct MIME types and SPA fallback.
- 05
Validate
Screenshots across three viewports, plus a report of console errors and 404s.
For builders, founders and agencies who need to rescue a site whose source is lost, rebrand a landing page fast, or migrate a deployed bundle into clean, maintainable code.
Free during public beta
$20 regular — after beta
Thanks — your download is ready:
↓ Download site-mirror.skillNo payment — free while in beta. Paid release after launch.
Solertio Board
An adversarial board of directors for your next decision. It critiques — it does not generate.
How it works
- 01
Submit
Describe the decision you’re about to make — a launch, a price, a pivot, a hire.
- 02
Convene
Three adversarial seats attack it independently: a hard-nosed CFO, a Legal/Compliance counsel, and a Chaos seat hunting for the failure mode you didn’t model.
- 03
Verdict
A board Chair synthesizes the objections into a single memo with one verdict:
GO,FIX-FIRSTorKILL.
For founders and operators about to commit money or months. Bilingual PL/EN. Bring your own Gemini key — the key and your submission never leave the browser, nothing is stored.
Free with your own API key
Convene the board →
BYOK: you paste a Gemini API key at runtime — it goes straight from your browser to Google, never to us.
Redaktor naczelny
It checks Polish copy before you publish it. Whether a text passes is decided by a program, not by the model’s opinion.
What you get
You buy once and get two packages: one for Claude Code, one for ChatGPT. Inside they are the same: a program that measures the text, a list of AI tells, rhythm rules and a dictionary of calques from English.
The program checks whether sentences and paragraphs vary in length, and separately hunts for the mistakes that disqualify a text: worn-out phrases, calques from English and rule-of-three lists made up on the spot. At the end it gives one of three answers: passed, did not pass, or not measured. The model may not declare a text good until it sees the word passed.
You build the brand voice yourself.
The wizard asks about the brand, the reader and the tone, then asks for three to five of your older texts and works out from them how you actually write. The package carries neither the author’s brand voices nor his quotation corpus.
After installing, you run one test. A broken paragraph has to come back as did not pass, the same paragraph after repair has to come back as passed, and only that pair proves the measurement works on your machine. A program that merely runs without an error proves nothing.
How the versions differ
The table below lists the differences without softening them. One matters when you buy: in ChatGPT the measurement runs in Code Interpreter, so without a plan that includes it there is no measurement, and the product loses the thing that matters most in it.
| Capability | Claude Code | ChatGPT |
|---|---|---|
| Rhythm and hard-violation measurement | yes | yes, via Code Interpreter |
| Second linter (Node) | yes | no — replaced by a mandatory read-through, noted as single-tool measurement |
| Voice wizard writes the files | yes | generates them, you upload them by hand |
| Release integrity check | yes | no — on update you replace the whole set of files |
| Repository files and batch work | yes | limited to chat files |
The Claude Code version needs Python 3 and Node. You get two zips with text and code, with no technical locks and no expiry date, but the download links live for only 30 days, so fetch the packages right away and keep them yourself.
149 zł
gross, incl. 23% VAT · launch price, 199 zł later