Dmytro Bondarets at his Tallinn workbench, holding a printed AI-tools catalogue in one hand, the violet bench light the only colour in the frame

The bench in Tallinn

Who explains these AI apps for you?

This site is run by Dmytro Bondarets, based in Tallinn, Estonia. One reader-focused notebook on AI assistants, research tools, coding helpers and media apps — no agency team, no newsroom desk, no affiliate desk behind it.

Umera tn 2-152, Lasnamäe, Tallinn — correspondence in English, Monday to Friday, 9:00 AM – 6:00 PM.

The point of the site

One notebook, updated as tools change.

Practical English-language guidance on AI apps and artificial intelligence solutions, written for individuals and small teams working in Estonia. If you have ever opened a dozen tools in one afternoon and finished none of the work, this site is aimed at the gap between the promise and the boring reality of a task you still have to deliver.

Pages describe the workflow first and the buttons second, because the menus change and the work does not. When a screen moves, the note still says which stage of the job the tool was doing. When a limit changes, the note says which small team it affected and what they had to do instead.

Coverage

What Dmytro covers.

Eight types of work get proper workflow notes. Everything else gets left alone — if a tool does not help a small team finish work, it does not get a long page.

  1. 01

    Drafting with Claude-style assistants

    Outlines, tone passes, structured first drafts a human then edits — the notes record where the assistant saved an hour and where the edit cost more than writing from scratch.

  2. 02

    Sourced research

    Finding answers with citations you can open yourself. The note each time is whether the tool shows its sources and how easy it is to check them.

  3. 03

    Knowledge bases with Notion AI

    Turning scattered team pages into something a new teammate can actually ask. Coverage looks at how much structure has to exist before the answers get useful.

  4. 04

    Polishing with Grammarly

    The last pass over client work written by a non-native speaker. The notes keep an eye on which suggestions help and which quietly flatten a sentence that was already fine.

  5. 05

    HR workflows with Workday

    Where AI sits in hiring and internal process — screening notes, job descriptions, onboarding material — and where a person should still be the one reading the material.

  6. 06

    Prototyping with Google AI Studio

    Getting a rough idea in front of a client or a colleague in an afternoon, before anyone commits to a build. Notes cover how much can be shown and what to strip before sharing.

  7. 07

    Media production with Runway and Descript

    Short-form video and transcript-based editing for people who shoot once and cannot reshoot. Fast-turn production is what the notes actually test.

  8. 08

    Building with Tabnine and Lovable

    Code completion and app building for people who know roughly what they want to ship. Notes cover what a junior developer, or a non-developer, gets out of each.

View from a Tallinn apartment window: a snowy Lasnamäe courtyard outside, and the reflection of a desk with a browser open inside

Why Estonia, why English

Two languages, one working day.

Many readers are students, freelancers and small teams working in English while living in Estonia. University material, client work and startup documentation routinely cross between languages, and the tooling has to hold up in both directions.

Guidance here stays in plain English and stays practical about the parts that actually trip people up: file formats, sharing permissions, what happens to a document when it leaves your laptop, and how to hand a result to a colleague who was not on the call. The aim is usable advice for Ulemiste City freelancers as much as for Tartu students writing their first English-language report.

It also shapes what gets covered here rather than elsewhere — a tool that works beautifully only inside one language or one corporate account is a smaller story for a reader in Tallinn than a tool that survives a handover.

A plain working method

Tested on ordinary tasks.

Dmytro tests tools on ordinary jobs that turn up in a small team's week: summarise a long PDF, draft an outline, clean a transcript, complete a function, generate a short clip. Not a benchmark suite, not a leaderboard — a set of tasks a freelancer or a small office actually runs.

Each test watches three things: whether sources are shown, whether the export works without a fight, and whether a teammate could follow the result without asking what happened. What survives that gets written up in plain language.

There are no lab scores on this site, only workflow observations you can repeat on your own machine. When the evidence is thin, the page says so.

Overhead view of a desk laid out as a working method: laptop with an outline document, notebook grid, sticky notes, pencil and a phone recording audio
The third of the method that never appears in a launch announcement: check the export, check the sources, check whether a teammate can follow.

How readers use the site

Most visits start with a task.

Nobody arrives wanting a tour of AI. They arrive with something due. The site is arranged so that a task can be answered in one page — and browsed at leisure when there is no deadline.

The task

Start with the job to be done

A marketer needs blog drafts. A junior developer needs code completion. A teacher needs transcript-based editing for a class video. The home page routes by the job, so the first click already answers the question you arrived with.

The comparison

See what fits your week

Comparison pages show fit — which tool suits a solo worker, which suits a two-person team, which one asks for a setup nobody in a small office has time to build. Fit is written as a condition, not a scoreboard.

The method

Check how the conclusion was reached

Method notes explain what was tested, on what kind of task, and what the limit was. You can read one page and act, or browse the full set on a quiet afternoon and see where two tools disagree.

The kitchen test

Every claim answers three questions.

A tool only earns a page on this site when it survives a repeatable test: does the export survive, does the source actually show up, and can a teammate who was not in the room follow the result without a phone call. What is written here is what survived that filter.

  1. 01

    Does the export survive?

    A generated outline that cannot be opened in the tool your client uses is a dead end. The note records what came out, in what format, and what had to be fixed by hand.

  2. 02

    Does the source show up?

    For research, the note says whether the assistant named where an answer came from, and whether that source could be opened and checked. Where the answer is simply unsourced, the note says that too.

  3. 03

    Can a teammate follow?

    A result that only its author understands is not a workflow, it is a personal shortcut. The note asks whether the handover survives without a phone call and a re-explanation.

Conclusions are written in plain language and revisited when a tool changes a menu, a limit or an address. No lab scores are published here — the observations are the ones you can repeat at home.

What this site will not do

The line the notebook holds.

  • It will not rank a single best app for everyone. Fit depends on the work in front of you, and a ranking pretends that one answer suits a freelancer, a school and an office the same way.

  • It will not publish leaked prices or affiliate pushes. Where a plan has moved, the note says the figure is out of date rather than inventing a fresh one to fill a table.

  • It will not invent user reviews or star ratings. There are no five-star widgets here, because the only honest reviews would be ones people actually wrote.

  • When evidence is thin, the page says so — that is the point of running the notebook in public.

Keeping guidance current

Pages note the workflow, not the button.

AI apps change menus monthly. A page that only names buttons goes stale the moment a redesign lands, so the notes describe the workflow — what stage of a job a tool is doing, what it produced, what had to be fixed before the work went out. When Runway or Lovable shifts its editor, the description of fast-turn production still holds.

Updates focus on the behaviour that changes your output, not on every minor release. When a change is cosmetic, the note stays as it is. When it changes what comes out of the tool, the page gets rewritten and the date on the note moves.

The Tallinn notebook, on the bench since the first note. Pages are revisited whenever a tool changes what it does — reachable at [email protected].