
Learn data and AI skills by doing. Interactive courses, projects and certifications, with the #1 AI tutor for AI and data.
New York
Joined July 2013
- Tweets12.9K
- Following556
- Followers109K
- Likes8.8K
One-to-one teaching was always the best way to learn. It just never scaled.
Our CEO, @CornelissenJo, and Chief AI Officer, @yusufmsaber, joined @chain_ofthought Podcast’s @ConorBronsdon to break down what makes the AI tutor’s adaptive learning so effective, and why it costs up to 20 times more to run than a standard course.
They lifted the lid on:
- The A/B testing the AI tutor vs. original DataCamp courses
- 30 core teaching behaviors, taken from what great teachers do
- Why the AI tutor has to respond in milliseconds
- What companies miss out on by not hiring juniors
Replying to @CornelissenJo
@CornelissenJo: @DataCamp's AI tutor beat its classic courses on willingness to pay, engagement and satisfaction. It also cost 10 to 20x more per hour of learning.
Shipping the better product made the business worse on paper(initially) - here's how they're solving it, and why they think its worthwhile.
DataCamp co-founder and CEO Jonathan Cornelissen and Chief AI Officer Yusuf Saber join @chain_ofthought to walk through how the tutor works and take us through best practices for learning with AI:
- About 30 core teaching behaviors, each checked by tests that run several times over
- Sonnet doing about 70% of the work, with cheaper tiers starting to take more
- Why tool calling is mostly off the table at millisecond latency
- Pre-generating the next response while the learner thinks, and cache hit rates in the mid-90s
- Why open models kept losing on speed, hosting and caching
& more
Chapters:
0:00 Better product, worse economics
0:59 How the tutor helps a struggling learner
7:08 From static courses to AI tutoring
9:25 Multi-agent architecture behind the tutor
13:09 Scaling costs and testing open models
18:30 Economics of cost per learning hour
22:13 What great looks like for DataCamp
25:32 Latency budgets and precomputed lesson plans
33:44 Measuring learning outcomes with Bloom's research
40:11 AI skills gap for early-career workers
43:20 Evals built on tutor behavior expectations
50:10 Closing thoughts on how skills are changing
Brought to you by:
• @coralbricks_ai: lightning fast inference with Deepseek v4.1 Flash - get $10 in credits to try: coralbricks.ai/r/Z6V2RE
• @WalrusProtocol - Walrus Memory gives AI agents portable, verifiable memory that carries context across apps, sessions, and other agents: walrus.xyz/cot
• @inngest - durable execution for agents in production. Failed steps retry; completed steps are saved and skipped: inngest.link/cot-x
• @wisdomai_inc - In their latest benchmark, Claude Code alone topped out at 73% accuracy. WisdomAI closed the gap, with accurate answers 3x faster at about a third of the cost: wisdom.ai/benchmark
Look up the @chain_ofthought podcast on YouTube, Spotify, Apple Podcasts (links below), or wherever you listen. Thanks, Jonathan and Yusuf!
We’re giving away a hand-signed Picasso print for the best AI-generated video that teaches one idea about AI or data in 60 seconds or less.
Before we name the winner on Wednesday, meet the 4 runners-up.
#WinAPicasso
Listen to the full episode: datacamp.com/podcast/sales-a…
1
333
Celebrate Diwali with 60% off, plus 6 months of DataLab, free: datacamp.com/promo/biggest-a…
1
320
The full story: datacamp.com/business/custom…
347








