Run a meeting poll, arrange bookings, organise training or prepare for a difficult conversation. I’m Eduard Ponomarev: after 15 years in HR automation, I’m building these tools with AI and sharing them as open experiments.
The Live polls and Academy workspaces and their participant pages are currently in Russian, and the games Bottleneck and Own Course are in Russian only.
Tools for meetings and training
Live polls, meeting booking and training management. These tools use no AI inside.
Hear the whole group during training, a retrospective or a working meeting, and agree the next step.
Key decisions and human checkpoints
By default, the facilitator reviews every answer and decides when to show the results, and the group sees totals only after three accepted answers. Participants answer without giving a name.
What AI did and what I checked
AI wrote the code and interface; I set the rules and checked the result. The service itself uses no AI: word clouds and totals follow transparent rules.
Limitations
A word cloud shows frequent words, not what they mean: the group draws the conclusions. A before-and-after comparison does not prove that training worked. The participant link stays valid for seven days.
Agree a time for an interview, a 1:1, an HR consultation or a feedback conversation without a long email thread. Guests choose an available time and book without creating an account.
Key decisions and human checkpoints
You set the working hours, date exceptions, meeting lengths, breaks and the deadline up to which guests can cancel or reschedule on their own. After that deadline, changes go through you.
What AI did and what I checked
AI wrote the code and interface; I set the rules and checked the result. The service itself uses no AI: available times and overlapping meetings follow predefined rules.
Limitations
There is no sync with external calendars: guests add the meeting to their calendar with an ICS file. This is a demonstration service: try it with fictional guests, not real candidates or employees.
Organise training: the programme, instructors, cohort schedules, participant registration and feedback after the sessions.
Key decisions and human checkpoints
You set the number of places and choose how registrations are confirmed: automatically or with the organiser’s approval. A freed place goes to the first approved participant on the waiting list. The organiser records attendance.
What AI did and what I checked
AI wrote the code and interface; I set the rules and checked the result. The service itself uses no AI: places, the queue and rating totals follow predefined rules.
Limitations
Combined participant ratings appear only after three responses. There is no sync with external calendars: participants add the schedule to their calendar with an ICS file. This is a demonstration service: try it with fictional participants, not real employees.
Practice
Rehearse a difficult HR conversation with an AI counterpart and get a review based on your own replies.
Prepare for a difficult conversation: feedback, workload, competing priorities or an AI pilot. The library has six fictional situations, and you can write your own.
Key decisions and human checkpoints
You choose the situation and the mode: role-play or coach. The review draws only on your own replies and the scenario’s criteria, with no scores and no judgement of your personality. What to say in the real conversation is up to you.
What AI did and what I checked
AI wrote the code and interface, and in use it plays the other person and drafts the review. I set the architecture and checked the result. The server, not the model, inserts the quotes in the review, taken verbatim from your replies.
Limitations
AI feedback can be wrong. Use fictional situations only. A voice attempt is shorter than a text one: up to 3 minutes and 6 replies.
People and AI on one team, a new manager’s first weeks and personal choices. Bottleneck, Own Course and the Bay’s first shift start straight away without creating an account.
A short strategy game about AI on the team. As the coordinator, you decide what goes to people and what to AI assistants, what to teach by example and when to step in yourself. The first shift is open to everyone.
Six game weeks of a new manager: all of the team’s work goes through you. A 12–15 minute session you can play straight away without signing in. The game is in Russian only.
Try management decisions in everyday situations: delegate a task, adapt your approach to a colleague, set priorities, limit work in progress and agree who is responsible.
What you decide
Where to direct the flow of work, which style this person needs for this task, what to put off, how much to take on and how far to delegate. On Friday there is a one-to-one and feedback. The game does not assess you as a manager.
How it was made
Claude (Anthropic) designed and built the game at my request: the mechanics, code, graphics and sound. No AI is called during play. The mechanics draw on books and research on management, from Linda Hill and Andy Grove to Little’s law; the sources and caveats are on the “About the game” page.
Limitations
Numbers and speeds in the game are game-design conventions, not research data. All people and situations are fictional. Progress is stored only in your browser; the game sends nothing to the server.
Look at your own choices from the outside: you plot a course, take on or skip optional errands and compare your forecast with how the voyage actually went.
What you decide
You choose the course, the cargo and the conditions; the line on the map does not steer the boat for you. The personal journal is optional: an initial assumption, an observation and a small test outside the game. The game does not assess your personality or draw psychological conclusions.
How it was made
GPT (OpenAI) created the scenario, code and illustrations on my brief. Claude (Anthropic) improved the game after review. No AI is called during play: the character’s lines are written in advance.
Limitations
The coaching frame draws on ICF, EMCC and the GROW model, but it is the author’s interpretation, not an accreditation. The game will never replace a coach, and its effectiveness has not been studied. Personal journal entries are not sent to the server: the journal lives only in the browser tab and is lost on reload unless you download it yourself.
How to get access
Register, choose the tools you need and briefly describe your task. I review each request personally. You will have your own workspace and can invite colleagues with suitable roles.
What happens next: an email to confirm your address → I review the request personally → an email with the decision.
Poll participants and guests booking a meeting or training place can use the shared link without creating an account. When booking a meeting or training, the organiser receives the name and contact details provided.