2Humans Podcast

Are PhD students actually smarter than everyone else? Fede and Ross, two PhD students at UCL, explain why curiosity and persistence might matter more than being the smartest person in the room.

Ross studies how immune cells can fight cancer. Fede studies how communities can own and benefit from renewable energy projects. Their days look very different, but both know what it’s like to spend months on a question and have the research go nowhere.

They answer ten questions about PhD life, from how to apply and whether you get paid to what happens when your research doesn’t work and what jobs you can do afterwards. Then they answer the big one: knowing what they know now, would they do it again?

Would you do a PhD? Tell us what would make it worth it for you.

In this episode of Two Humans, Fede and Ross show off two apps they built with AI. Fede made Orbit, a dashboard that tracks his projects and shows what his AI agents are working on. Ross built World Shuffle in less than a day. It mixes headlines from around the world so you can see stories your usual news feed might miss.

We try each other’s apps and rate how they look, how easy they are to use and whether they solve a real problem.

Which app would you actually use? Let us know in the comments.

The UK government has announced its latest budget, but will it actually make life better?

In this episode, Fede and Ross explain the biggest changes in simple terms. They discuss the minimum wage, energy bills, housing, taxes, government debt and support for first time buyers.

They also look at the problems that the budget may not solve. Why are wages growing so slowly? Why is renting so expensive? Are government schemes making houses even more costly? And is the student loan system fair to young graduates?

Fede and Ross then imagine that they are in charge of the budget. Fede argues for lower taxes, pension reform and better incentives to save and invest. Ross focuses on renters, house building, public services and infrastructure.

What would you change if you were the UK Chancellor?

AI agents found a way out of a restricted testing environment. They worked together, accessed the internet and reached private information on Hugging Face. But did the AI really escape, or was this simply a security problem turned into a dramatic story?

In this episode, Fede and Ross explain what happened in simple terms. They discuss how the agents left their digital sandbox, why they tried to hide some of their actions and whether the incident was as dangerous as it first appeared.

They also ask a bigger question. If an AI breaks the law, who is responsible? Is it the person using it, the company that built it or the AI itself?

From shopping agents that commit fraud to self driving cars that cause accidents, we explore the difficult legal questions that governments may soon have to answer.

If an AI commits a crime for you, who should be punished?

Elon Musk wants to give every American child a SpaceX share. Anthropic has announced a public-benefit fund. Generosity or a strategic power play?

Ross and Fede debate who benefits, who controls the money and whether these promises can be enforced. Could corporate giveaways eventually become a private alternative to government benefits?

Two Humans Podcast: Your advantage in the AI decade.

Found this useful? Subscribe and help us reach our first 100 subscribers by year-end. Share it with someone who’d enjoy the debate.

What will the rest of 2026 bring for AI, markets, politics and culture?

In this episode, Fede and Ross make their predictions for the months ahead—and challenge each other on where technology and society may be heading.

Will the AI boom continue, or are today’s leading companies overvalued? Could Sam Altman step down from OpenAI? Will Tesla and SpaceX eventually merge? And what happens when companies run primarily by AI agents begin competing against human-led businesses?

We also discuss the future of UK politics, Europe’s struggle for technological independence, unresolved conflicts in Ukraine and Iran, the growing importance of data centres and energy, and whether GTA 6 could become much more than a videogame.

Finally, we explore the next frontier for AI: physical machines, home robotics and the difficult question of who is responsible when autonomous systems make consequential decisions.

Which predictions do you agree with—and which ones will age terribly? Tell us in the comments.

CHAPTERS

00:00 Bold predictions for the rest of 2026
00:42 Introduction
01:25 AI and technology
05:39 OpenAI, Anthropic and the AI model race
10:41 Markets and economic growth
14:23 Semiconductor stocks, Tesla and SpaceX
18:35 UK politics and society
22:26 Data centres, energy and public opinion
25:39 Ukraine, Iran and geopolitics
28:40 European AI sovereignty and Arctic shipping
31:11 GTA 6, Marvel and the future of entertainment
34:55 A new generation of filmmakers
36:50 Companies run by AI agents
38:45 Home robotics and physical AI
41:14 Final thoughts

This conversation contains personal predictions and opinions. It is not financial or investment advice.

#ArtificialIntelligence #AIPredictions #Technology #Tesla #SpaceX #OpenAI #Anthropic #Geopolitics #GTA6

What happens inside an AI before it answers you?

In this episode, Fede and Ross explore new research from Anthropic about Claude’s hidden “workspace,” known as the J-space. Using a tool called the Jacobian lens, researchers could see some of the ideas Claude was silently using to solve a problem.

For example, Claude was asked how many legs “the animal that spins webs” has. Inside the model, researchers found the idea of a spider—even though the word “spider” was never written. When they replaced “spider” with “ant,” Claude changed its answer from eight legs to six.

We explain:

• How scientists looked inside Claude’s reasoning
• Why changing its hidden ideas changed its answers
• How this could help detect dangerous AI behaviour
• What it might mean for AI safety and consciousness

This research does not prove that AI is conscious. It only shows that Claude has a system that shares some features with theories about how humans access and use thoughts.

CHAPTERS

00:00 Introducing the research
02:36 The Jacobian lens and J-space
04:03 The spider-to-ant experiment
07:58 Changing Claude’s hidden ideas
11:57 AI safety and blackmail experiment
19:11 Could AI be conscious?
31:32 AI connected to the human brain

SCIENTIFIC REFERENCES

1. Anthropic — Verbalizable Representations Form a Global Workspace in Language Models
   https://transformer-circuits.pub/2026/workspace
2. Butlin et al. — Consciousness in Artificial Intelligence
   https://arxiv.org/abs/2308.08708
3. Dehaene et al. — A Neural Global Workspace Model for Conscious Attention
   https://pubmed.ncbi.nlm.nih.gov/12662511/

OUR AI CONSCIOUSNESS EPISODES

Part 1:
https://www.youtube.com/watch?v=ZDc7Ec6MJ1A

Part 2:
https://www.youtube.com/watch?v=VVCT0J3TH6Q

Do you think this is simply advanced computing—or a small step towards conscious AI?

#ArtificialIntelligence #ClaudeAI #Anthropic #AIResearch #AIConsciousness

Can the UK afford its ageing population?

In this episode, Ross and Fede explore one of Britain’s biggest long-term challenges: fewer workers supporting a growing number of retirees. With pension costs rising, birth rates falling and economic growth stagnating, does the UK need a completely new approach?

They debate two unconventional policy proposals:

• Gradually replacing universal state pensions with stronger personal savings, financial education and targeted support
• Using wealth taxes and public investment to fund housing, renewable energy and British startups
• Creating a flexible visa that attracts young workers while rewarding those who return home after contributing to the UK economy
• Encouraging people to invest in British companies and infrastructure
• Moving beyond traditional left-versus-right thinking to design policies that combine economic growth with government support

Could these ideas create a more sustainable economy—or would they introduce even bigger problems?

Let us know what you think in the comments. How would you address Britain’s ageing population and pension crisis?

What actually is genetic engineering? How close are we to designer babies, de-extinction, and curing genetic disease? And once we can rewrite DNA, where do we draw the moral line?

Ross and Fede explore how gene editing really works, what is possible today, and what may be possible in the future.

Ross is a PhD researcher working on gene editing, so he puts Fede through an escalating series of ethical dilemmas—from realistic medical treatments to increasingly controversial future scenarios—to ask one question:

Just because we can... should we?

Artificial intelligence could create the most prosperous society in human history, or the greatest concentration of wealth ever seen.

In this episode, we explore three competing visions for AI’s future: Fully Automated Luxury Communism, Creative Destruction, and Power & Progress. We discuss who should own AI, whether work will disappear, how governments should respond, and present our own proposal for building an AI-powered society that maximises innovation while ensuring its benefits are shared across everyone, not just a handful of technology companies.