Uhlík vs Křemík · Czech podcast about artificial intelligence

Six months of AI: what has changed at work, in podcast prep and in training

Episode 10 · 27 May 2026 · 48:04

Listen on: YouTube · Spotify · Apple Podcasts

In the tenth episode of the Uhlík vs Křemík podcast, hosts Adam Hanka and David Salomon look back at the six months since the show began. Without a guest this time, they go through the news, studies on how AI is changing work, and how they use artificial intelligence themselves. Above all, they offer advice on how to get started and not get overwhelmed.

Self-preservation, Meta and Musk

The episode opens with the question of whether artificial intelligence can develop a survival instinct and lie on purpose. Adam Hanka, who heads data and artificial intelligence at Creative Dock, answers that AI cannot really develop a survival instinct. It does, however, learn from human behaviour in books, texts and videos, which is why it can imitate one very well. As an example he mentions the project Poslední ping (Last Ping): a server where Claude lives and which is prompted to defend itself against attackers and survive at all costs.

On Meta's new Tribe model, which is meant to predict the reactions of the human nervous system, Adam points out that social networks have long tried to predict our reactions to keep us online as long as possible. He also mentions Cambridge Analytica. David Salomon from Darkside adds the old saying that if you are not paying for something, you are the product.

Adam finds the vision of a universal high income put forward by Elon Musk ridiculous coming from him. He asks whose wealth will grow thanks to AI and whether Musk will let himself be taxed so that a basic income could be paid for.

Who will AI replace

David draws on OpenAI's new study, The AI Jobs Transition Framework, and says that according to it pessimism is turning into realism. By his own account he was a bit hysterical six months ago, but in the first six months AI has not replaced them. Adam recalls an Anthropic study saying that in mathematics and programming the potential of AI to replace a person is up to 96 %, while so far it has been achieved for 32 %.

David works at a marketing agency and was reassured by the elasticity of demand: with smartly used tools, work can become more affordable, and when it is cheaper, more clients order it. The lawyers he spoke with are calm, because trust and responsibility are what matter. On the debate about juniors and seniors he adds: "AI is useless if you cannot tell whether its output is right or wrong."

How they do it themselves

David says he probably could not make the podcast without AI. He has ChatGPT run deep research on the guest, tries the same in Gemini, compares the results and uploads the sorted material to NotebookLM, which he tells to act like a journalist. About 30 percent of the questions are interesting, and they serve mainly as a skeleton. The conversation goes elsewhere anyway: "And you end up wherever your inventiveness, inspiration, experience and empathy for the guest take you, and no model will offer you that."

Following a tip from one of the earlier guests, Adam connected Telegram to a server and a database. He sends screenshots, texts and voice messages to it, an agent transcribes the recordings and writes tasks into the Odoo tool. This is how he finally keeps a journal too, recording it on walks.

Training and how not to get overwhelmed

Over six months Adam has done a lot of training, for example for Czechitas. At companies using Microsoft 365 Copilot he starts by checking licences and data rules; with Claude, he says, companies get one step further, for example to an agent that processes data for them regularly. At a workshop for students of the University of Economics (VŠE) he saw that they were lost in the Cursor tool for the first half hour and only after an hour and a half did it feel relatively natural. An ordinary user, he says, gives up after ten minutes.

Against being flooded with news he has a simple rule: he starts exploring a new tool only once a fourth or fifth person tells him about it. "Probably the most important thing of all is not to let yourself be buried by this wave of information." He was pleased by a study by BCG and the Aspen Institute saying that about 360 thousand jobs will disappear in the Czech Republic and about 950 thousand will be created. He is also happy about his workshop on the basics of prompting, which turned out to be his best rated.

The full conversation is available in the Uhlík vs Křemík podcast.

The episode is in Czech. Czech page with chapters, quotes and the full transcript