Skip to content

2026

  • 3 min read

How to Document an AI Wednesday Session

AI Wednesday is a weekly, community driven gathering at TinkerSpace where people connect, share, and build.

From 2026 onwards, every session should have a short journal entry, usually written by the host.

Why we document AI Wednesday

AI Wednesday is a weekly community gathering, but the conversations, demos, and learnings should not end when the session ends.

By documenting each AI Wednesday, we:

  • Keep a clear log of what was discussed and built
  • Help members who could not attend stay updated
  • Make it easy for new members to understand the community’s direction
  • Create a shared reference for ideas, projects, and follow ups
  • Build a long term archive of our community journey

These short journal entries are not meant to be detailed reports. They are lightweight snapshots of each session, written so that any community member can quickly read, catch up, and continue the conversation.

Documentation steps

  1. Set up MkDocs Material on your computer
  2. Run the site locally
  3. Add a new journal post using the template
  4. Push changes and raise a pull request

Setup on your computer

First, you need setup the computer to run the mkdocs locally.

Prerequisites
  • Install Python and pip - follow the guide here
  • Install MkDocs - follow the guide here.
  • Install Git - follow the guide here

Fork, clone, and run locally

  1. Fork the repository

  2. Open the upstream repository: tinkerhub/AIWednesday

  3. Click Fork (creates your copy under your GitHub account)

  4. Clone your fork

git clone https://github.com/tinkerhub/AIWednesday.git
cd AIWednesday
  1. Start the local dev server

Run this from the folder that contains mkdocs.yml:

mkdocs serve

MkDocs will serve the site locally (usually on http://127.0.0.1:8000/)

Create a new AI Wednesday journal entry

The site journal is powered by the Material blog feature (you can see it live at /blog/)

  1. Create a new branch
git checkout -b add-AIWednesday-journal-YYYY-MM-DD
  1. Add a new post file

Follow the existing pattern in the repository for where posts live.

  • docs/blog/posts/YYYY-MM-DD-title.md

  • Add front matter

Add metadata for your Journal post,

---
title: 'AI Wednesday XX : Name for the AI Wednesday'
date: YYYY-MM-DD
authors: [autor]
slug: ai-wednesday-xx
description: >
  A Small Description
---

If this is your first time contributing, please add your details to the authors file before submitting your post.

open the file docs/blog/authors.yml and add new entry

  author-username:
    name: Author Full Name
    description: Author tag line
    avatar: https://avatars.githubusercontent.com/u/xxxxxxxx
    url: Tinkerhub App profile URL

In the avatar section, replace the xxxxxxx to your github profile id.

Journal template (copy paste)

Use this template inside your post:

---
title: 'AI Wednesday XX : Name for the AI Wednesday'
date: YYYY-MM-DD
authors: [autor]
slug: ai-wednesday-xx
description: >
  A Small Description
---

## Overview
Write 3 to 6 lines on what happened this week. Mention the theme and the general vibe.

## Topics
- Topic 1
- Topic 2
- Topic 3

## Project Presentation
- Name – project title (one line summary if needed)
- Name – project title

## Photos
### Group photo
![Group photo](../assets/XX/group-photo.jpg)

### Activity photo
![Activity photo](../assets/XX/activity-photo.jpg)

## Highlights
- One key takeaway (learning, decision, win, or community moment)

## Next Week
- Topic: TBD
- Host: TBD
Notes for photos:
  • Keep filenames simple and consistent.
  • Store images in the same place the repo already uses for images.
  • Always add meaningful alt text.

Preview your changes

While mkdocs serve is running, open the local site and check:

  • The post appears in the journal list.
  • Images load.
  • Headings look correct and spacing is clean.

Commit, push, and open a pull request

  1. Commit and push to your fork
git add .
git commit -m "Add AI Wednesday journal for YYYY-MM-DD"
git push -u origin add-AIWednesday-journal-YYYY-MM-DD
  1. Create a pull request to upstream

On GitHub, open your fork and you should see a prompt to create a PR. GitHub’s flow for PRs from forks is documented here

PR checklist:

  • Title includes the date and session name.
  • Post follows the minimum structure.
  • Photos included or clearly marked pending.
  • Previewed locally.

Suggested style (keep it consistent)

  • Prefer short paragraphs and bullet lists.
  • Avoid long intros.
  • Name people and projects clearly (credit matters).
  • If something is TBD, write “TBD” instead of leaving it blank.
  • 2 min read

AI Wednesday 02 : Under the Hood of Modern LLMs

Overview

This week’s AI Wednesday session was about understanding how modern AI models like ChatGPT actually work behind the scenes. Instead of treating AI as something magical or human-like, we broke it down into simpler ideas and mechanisms.

We started by clearing a common misconception that these models “think” or “understand” things. The discussion focused on how LLMs are trained to predict the next piece of text, and how doing this at a very large scale creates the illusion of intelligence.

The session was mostly conceptual and discussion-driven. The goal was not to teach people how to build models, but to help everyone develop a clear mental model of what is happening inside these systems.

Topics

  • Why LLMs do not think like humans and what they actually do instead
  • What “next token prediction” means in simple terms
  • How text is broken into tokens and converted into numbers
  • What embeddings are and how they represent meaning
  • A brief look at how language models evolved from simple word prediction to transformers
  • What attention is and why transformers made modern AI possible
  • Some important limitations of LLMs and why they sometimes fail

Photos

img1

Group photo

Highlights

  • A major takeaway was realizing that modern AI feels intelligent not because it understands, but because it has learned patterns from huge amounts of human language.
  • 3 min read

AI-Wednesday 01: Creating the 2026 plan for AI Wednesday while reflecting on the past year activities and improvement areas

2025 Statistics

AI Wednesday saw a successful run in 2025, marking a huge jump in both the consistency and quality of sessions taken. Among the 53 total sessions taken, 31 of them were taken this year with around 188 unique attendees and 12 Organisers

Though consistency during the later half of 2025 has improved, the major challenge remains the less number of hosts to take the sessions. The focus on the topics for the sessions have also switched to a much detailed approach leading to a higher involvement of more serious participants.

Summary report

2025 marked a turning point for AI Wednesday - not just in numbers, but in purpose. What began as a space to casually exchange updates on AI trends gradually evolved into a focused learning-driven community. Early sessions leaned toward showcasing projects, discussing breakthroughs, and exploring newly released tools. These conversations helped build momentum and attracted a diverse group of curious minds, but as the year progressed, a deeper realization emerged: attendees weren’t just looking to know what’s new - they wanted to understand how things work.

This shift shaped the second half of the year. Sessions became more structured, topic-driven, and intentional. Instead of broad overviews, each AI Wednesday began centering around a single concept that participants could meaningfully take away - something they could later explore, build upon, or apply. Knowledge became the north star. As a result, engagement deepened, discussions became more thoughtful, and participants walked away feeling that their time had genuinely added value to their learning journey.

2025 Statistics

Complex and high-impact topics such as Agentic AI and MCP were explored in depth, sometimes spanning multiple sessions to do justice to the subject. This approach not only improved conceptual clarity but also encouraged continuity - participants returned week after week to build on what they had learned earlier. The community itself grew stronger through this consistency, drawing in a wide demographic ranging from school students to experienced working professionals, all united by a shared intent: to learn AI seriously and meaningfully.

By the end of 2025, AI Wednesday had matured from a discussion forum into a learning ecosystem—one that values depth over hype, understanding over surface-level trends, and community-driven growth over passive consumption. This foundation now sets the stage for a more ambitious and impactful 2026.

Future plans

During the review and planning discussion, we aligned on the following actions for the coming year:

  • Topic Planning: Topics should be expanded to include more in depth learning of the theoretical basics like machine Learning and Deep Learning, rather than sticking to mostly Generative AI.
  • Documentation: From 2026 onwards, every AI Wednesday will be documented. The host is responsible for a short journal entry, following simple guidelines, giving visibility to both the host and the community.
  • Intro to AI: A pre-curated session would be held every month called Introduction to AI where beginners can get their initial exposure to what AI is and the basics of computing.
  • Practical Projects: Encourage a more practical approach to learning AI, with topics such as local hosting and inference and Edge AI.
  • Physical AI Collaboration: Work with Maker Thursday to explore Physical AI, focusing on AI running on hardware, while setting up a project that can be displayed on Tinkerspace.
  • Inference Guide: Explore and curate a guide that updates the cost of AI inference with suggestions on use cases for leading models.