DataTalks.Club

The episode has a handful of useful observations - context engineering, multi-agent team structures, using failure-rate comparisons as an AI metric - but they are heavily diluted by personal anecdotes, circular restatements, and vague platitudes. The ratio of novel ideas to filler is low for a 62-minute runtime.

“if you give all the right context and if you have everything lined up for AI agents or the model that you use, the quality is actually pretty good. But you need to invest a lot of time to build that context, to maintain it.”

“what's the failure rate of code generated by engineers versus failure rate of code generated by AI agents? That's a good metric to track.”

Most takes are widely circulating: AI creates more work, big companies pave the way, juniors are being squeezed out. The junior-vs-subscription cost comparison is mildly interesting, and context engineering gets a genuine nod, but neither is developed with enough depth to be truly contrarian or first-principles.

“it's cheaper to hire a junior engineer than run Claude or Codex, right? Because the costs are higher than what you would pay an actual human.”

“there's a leaderboard, all right, who spends more tokens, right? Uh, and then people just read stories online. People just run LLMs in recurse, do uh, some random stuff, right? Just to generate, to burn more tokens so that you look better on the leaderboard.”

Ivan is a genuine practitioner with a real technical background (transformer model thesis, multilingual NLP at scale, 5+ years in IAM engineering management), which gives him credibility. However, he is mid-tier in terms of seniority and scale - he frequently references what he 'hears from the industry' rather than sharing first-hand results from a significant organisation.

“My master thesis was on the Transformer model. I was doing some tweaks to it way back when it first released.”

“I've worked a lot in AI before. I mean, I, um. My master thesis was on the Transformer model.”

There are some named reference points - Amazon's December policy shift, Uber's token budget experiment, the 'Three Man Team' GitHub project, DORA metrics - but none are accompanied by actual numbers, timelines, or cited data. The examples remain anecdotal and unverified in the conversation.

“there's a project on GitHub called Three, uh, Man Team. And so one person is a developer agent, the other one is a reviewer agent and the third one is like a product agent.”

“Amazon, right? They had their problems in December where they tried something, you know, AI does everything and then it backfired to some extent and now they changed the policy. Now AI code needs to be reviewed by a senior engineer.”

The host occasionally asks substantive questions (measuring AI ROI, DORA metric applicability, org-level adoption frameworks) but repeatedly derails the conversation with extended personal anecdotes and answers his own questions in the setup. There is almost no pushback or genuine challenge to any of the guest's claims.

“How do we measure if it's useful or not? Because, um, there is a story probably also heard from Uber, right?”

“my son is like other kids, he's very inventive. So he has uh, a pair of pants and there was a hole on this and he fixed it with a scotch tape. Right. So he just put scotch tape on the pants and then it was the fix.”