Turing Post

In this episode of Inference, we ride through San Francisco – as one of the first to do this test drive – and talk about what’s changing in autonomous driving: cheaper hardware, better models, synthetic data, and a whole new approach to building the software behind the wheel. Ali Kani has been at NVIDIA Automotive for almost 8 years – he’s been through all the ups and downs, and he’s eager to share.

*We talk about:*

Why Level 2 is already possible with a surprisingly cheap sensor setup

What is still missing for Level 4

Why next year could matter for Level 4

How NVIDIA combines an end-to-end driving model with a classical safety stack

​​Why open source matters for the future of autonomous driving

Why synthetic data and simulation may matter as much as real-world driving data

How different cities, laws, and driving cultures change the way autonomous systems behave

Why the goal is bigger than one self-driving car – it’s making many cars autonomous by open sourcing the whole stack (it’s HUGE)

We also experience live what still makes urban driving hard: construction, cyclists, congestion, weird negotiations at stop signs, and all the messy little moments humans barely notice but cars have to handle perfectly.

What I liked about this conversation is that it makes the shift feel very real. *We’re moving from self-driving built inside closed labs to self-driving becoming a shared capability that can spread across the whole car industry.*

This is a conversation about a future that starts tomorrow. It’s open and very exciting.

Chapters:

0:00 The Future of Self-Driving Starts Now

0:19 Open Autonomous Driving Beyond Tesla and Waymo

1:07 Inside NVIDIA’s Low-Cost Level 2 Self-Driving Stack

1:48 From Level 2 to Level 4: Hyperion, Thor, and Redundancy

2:43 How NVIDIA Combines End-to-End AI with Safety Guardrails

3:56 What Changed in AlphaMaio Since GTC

5:12 The Key Technologies Needed to Solve Self-Driving

7:22 Real Data vs Synthetic Data in Autonomous Driving

9:21 Driving Through Real San Francisco Traffic

18:55 AlphaDream and the Next Generation of Simulation

*Follow on*: https://www.turingpost.com/

https://www.turingpost.com/p/av

*Did you like the episode? You know the drill:*

📌 Subscribe here and here (https://www.turingpost.com/subscribe) for more conversations with the builders shaping real-world AI.

💬 Leave a comment

👍 Like it

🫶 Thank you for watching and sharing!

*Guest:*

Ali Kani, Vice President and General Manager of Automotive, NVIDIA

https://www.linkedin.com/in/ali-kani-b22198

https://blogs.nvidia.com/blog/author/alikani/

Read more:

https://www.turingpost.com/p/selfdriving

https://thefocus.ai/posts/the-car-wash-test/

What does it mean to be a programmer when you are no longer typing most of the code? Which skills become more important in an agent-driven workflow? Will coding agents eventually take over most software implementation? And if that happens, what is left for the human engineer besides pushing prompts around like a confused project manager with Wi-Fi?

All of it and more in this part – watch it.

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:* Michael Bolin, tech lead on Codex, OpenAI

https://www.linkedin.com/in/michael-bolin-7632712/

https://x.com/bolinfest

https://github.com/openai/codex

Chapters:

0:00 — Do You Still Need to Learn Coding?

0:18 — From Systems to Humans: The Future of Programming

0:39 — Switching Mindset: Building for Agents vs Developers

1:13 — What Happens When Agents Consume the Web?

1:27 — Programmer Identity in the Age of AI

2:15 — Are Engineers Building More Than Ever?

2:37 — Key Skills for Engineers Working with AI Agents

3:59 — Will Agents Take Over Coding?

4:57 — Engineering Taste vs AI Decisions

5:10 — From Idea to Product Faster Than Ever

6:01 — Risks: Losing Human Judgment Too Early

6:42 — Do We Still Need Humans in the Loop?

8:06 — Book That Shaped a Builder’s Mindset

📰 Transcript:https://www.turingpost.com/p/bolincodex

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

#AI #OpenAI #Codex #MichaelBolin #SoftwareEngineering #Programming #CodingAgents #AIAgents #DeveloperTools #HarnessEngineering #FutureOfWork #Engineering #TuringPost

Stronger models clearly pushed Codex to new heights. But without the right harness around them, those models would not be able to operate reliably, and – most importantly – safely on a real developer’s machine. At least, not yet.

In this episode of Inference, I talk with Michael Bolin – lead for open source Codex at OpenAI – about the engineering layer that makes coding agents actually function: the agent loop, sandboxing, tool orchestration, and the design decisions that determine how much freedom an agent should have.

In this conversation, we get into:

What a harness actually is and why every coding agent needs one

Can a model be enough for a reliable coding workflow

Why do they build harness as small and tight as possible

How Codex handles sandboxing and security across OS

Why safety and security are not the same thing in agentic systems

How coding agents are changing the daily workflow of developers

Why documentation, tests, repo structure, and agents.md suddenly matter more

Whether too much context can make an agent worse

Why Michael believes the future may involve fewer tools, but more powerful ones

If you’re trying to understand where coding agents are actually going, this episode is for you.

Subscribe to the channel to be notified about Part 2, where we discuss what becomes of the software engineer in the age of agents.

Chapters:

0:00 The New Inner Loop of AI Coding Agents

0:17 Introduction: Michael Bolin and Open Source Codex

1:17 What the “Harness” Is in AI Coding Agents

2:13 Security and Sandboxing for AI Agents

4:33 Codex Launch and Rapid Growth

5:25 The Codex App: A New Interface for Developers

6:36 How Coding Agents Change Developer Workflows

10:04 Writing Codebases and Documentation for AI Agents

12:44 Context Engineering and Prompting for Codex

16:02 Model vs Harness: What Really Matters for Agents

19:23 Multi-Agent Systems, Tools, and the Future of AI Development

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe here and here (https://www.turingpost.com/subscribe) for more conversations with the builders shaping real-world AI.

💬 Leave a comment

👍 Like it

🫶 Thank you for watching and sharing!

*Guest:*

 Michael Bolin, tech lead on Codex, OpenAI

https://www.linkedin.com/in/michael-bolin-7632712/

https://x.com/bolinfest

https://github.com/openai/codex

📰 Transcript: https://www.turingpost.com/bolin1

*Turing Post* – AI stories from labs the Valley doesn't cover.

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

Tags: #AI #OpenAI #Codex #CodingAgents #DeveloperTools #AgenticAI #SoftwareEngineering #HarnessEngineering #Harness

And I’ll be honest: I left this conversation more skeptical than I expected. They raised $2 billion last year. But where the results?

Reflection’s thesis is huge – build the missing Western open base model, then use RL to push it to the frontier. The problem is that this is also the slowest path in the game. “All hands on deck building the model” means no clear wedge product yet, few concrete proof points, and a lot of execution risk while closed labs keep shipping.

Am I missing something? Watch the video and leave your opinion in the comments

Chapters:

0:00 Building AGI and the Mission Behind Reflection

0:25 From AlphaGo to Today: How AI Progress Really Happens

2:11 Breakthroughs vs. Engineering: What Still Matters Most

3:10 Defining AGI and Why It May Not Need Huge Breakthroughs

3:41 Why Reflection Shifted from Coding Agents to Frontier Models

5:15 The New Focus: Open Frontier Models and General Agents

6:33 Bottlenecks in Building Frontier AI: Team, Compute, and Scale

7:48 AI Tools, Internal Workflows, and Model-First Strategy

8:24 Can Open Models Catch Closed Labs?

10:34 Reinforcement Learning, Research Priorities, and Advice for Young Builders

14:01 Joining DeepMind, Open Science, and the Book That Shaped Him

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:*

Ioannis Antonoglou, Co-Founder, President & CTO at Reflection AI https://x.com/real_ioannis https://www.linkedin.com/in/ioannis-alexandros-antonoglou-45393253 https://reflection.ai/

📰 Transcript: https://www.turingpost.com/nathan

*Turing Post* – AI stories from labs the Valley doesn't cover.

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

Tags: #reflectionai #opensource #deepmind #ai #openclaw #aisafety

In Part 1, we talk about openness as an actual strategy: why open models can move faster, why “sovereignty” matters for enterprises and governments, and why safety might improve when the ecosystem can stress-test the system instead of guessing.

We also get into the uncomfortable part: capable open agents can misbehave in public, fast (OpenClaw is the recent reminder). Is that a reason to close everything up, or a reason to make the risks visible and fixable?

Topics covered:

 – Why a former DeepMind builder chose open weights

 – Open models as a commercial engine (and what investors bought)

 – Openness, safety, and “more eyes on the system”

 – Concentration of AI power in closed labs

 – Who open frontier models are really for (research, enterprises, governments)

Subscribe for Part 2: how Reflection plans to compete with closed labs and what they’re building under the hood.

Chapters:

0:00 — “No One Was Sharing This Information”

0:16 — From DeepMind to Reflection AI

0:52 — Why Move from Closed Labs to Open Weights?

2:20 — Pitching Open Models Before the DeepSeek Moment

3:31 — What Changed in the Past Year

4:43 — Why Openness Accelerates Scientific Progress

6:06 — Open Source vs Safety: The Open Claw Case

7:19 — The Real Concern: Concentration of AI Power

8:23 — The Open Source Paradox

9:11 — The Value Proposition of an Open Frontier Model

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:*

Ioannis Antonoglou, Co-Founder, President & CTO at Reflection AI https://x.com/real_ioannis https://www.linkedin.com/in/ioannis-alexandros-antonoglou-45393253/ https://reflection.ai/

📰 Transcript: https://www.turingpost.com/antonoglou_part1

*Turing Post* – AI stories from labs the Valley doesn't cover.

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

Tags: #reflectionai #opensource #deepmind #ai #openclaw #aisafety

Nathan Lambert doesn’t believe open models will ever catch up with closed ones, and he explains clearly why. But he also argues that this is the wrong framing.

Nathan is a research scientist at the Allen Institute for AI, the author of the RLHF Book, and the writer behind the Interconnects newsletter. He’s also one of the clearest voices on what open models are for, and just as importantly, what they are not.

We talk about how academic AI research lost influence as training scaled up, why open models became the main place where experimentation still happens, and why that role matters even when open models trail frontier systems. We also discuss why China’s open model ecosystem developed so differently from the US one, and what that tells us about incentives, talent, and access to resources.

From there, the conversation moves into the mechanics: post-training and reinforcement learning complexity, data availability, coding agents, hybrid architectures, and the very practical reasons most people continue to rely on closed models, even when they support openness in principle.

This is a conversation about how AI research actually moves, where open models fit into that picture, and what it means to build systems when the frontier is expensive, fast-moving, and increasingly product-driven.

This conversation offers a realistic look at where the open ecosystem stands today. Watch it!

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:*

Nathan Lambert, Research Scientist at Allen Institute for AI (AI2)

https://x.com/natolambert

https://www.linkedin.com/in/natolambert/

https://www.interconnects.ai/ (his newsletter on open models + RL + everything important in AI)

https://rlhfbook.com/ - The RLHF Book

https://allenai.org/

*Links:*

State of AI in 2026 (Lex Fridman interview): https://www.youtube.com/watch?v=EV7WhVT270Q&t=10206s

NVIDIA’s path to open models https://www.youtube.com/watch?v=Y3Vb6ecvfpU

OLMo models: https://allenai.org/olmo

NVIDIA Nemotron: https://developer.nvidia.com/nemotron

SpaceX + xAI partnership: https://www.spacex.com/updates#xai-joins-spacex

Season of the Witch (book): https://www.simonandschuster.com/books/Season-of-the-Witch/David-Talbot/9781439108246

📰 Transcript: https://www.turingpost.com/nathanlambert

*Turing Post* – AI stories from labs the Valley doesn't cover.

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

When MiniMax's RL training wouldn't converge, they debugged layer by layer until they found it: fp32 precision in the LM head. When their models learned to "hack" during training, exploiting loopholes to maximize rewards, they had to rethink alignment from scratch. When benchmarks said their models were good but production said otherwise, they discovered the problem: environment adaptation.

Olive talks about working at a pace where new models drop at midnight and you test them at midnight. How they use an internal AI agent to read every new paper published overnight. Why they sit with developers during experiments to catch dangerous behaviors in real-time. What "ICU in the morning, KTV at night" means when results swing wildly. How problem-solving becomes discovery when you're debugging behaviors no one has seen before.

This is how Chinese labs are moving fast: first-principles thinking, engineering discipline, and willingness to work whenever the model in experimentation requires you to.

We spoke on Sunday at 9 pm Beijing time. Olive was still waiting for results from new model experiments, so my first question was obvious: does everyone at the company work like this?

*Follow on*: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:* Olive Song, Senior Researcher at MiniMax

MiniMax: https://www.minimaxi.com/

Models: https://huggingface.co/MiniMaxAI

*Links:*

vLLM: https://github.com/vllm-project/vllm

SGLang: https://github.com/sgl-project/sglang

📰 Transcript: https://www.turingpost.com/olive

Chapters:

0:00 – Reinforcement Learning and Unexpected Model Behaviors

3:08 – Roleplay, Alignment, and “AI with Everyone”

4:02 – How AI Changes Daily Life and Productivity

4:59 – Inside Miniax: How Researchers and Engineers Work Together

5:32 – Human Alignment and Safety in Open Models

6:16 – Why Engineering Details Matter More Than Algorithms

8:17 – Open Weights: Benefits, Risks, and Responsibility

10:57 – Specialization vs General AI Models

12:07 – Agentic AI and Long-Horizon Tasks

29:50 – AGI, Creativity, and the Future of AI

*Turing Post* – AI stories from labs the Valley doesn't cover.

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

#MiniMax #ReinforcementLearning #AIResearch #OpenWeights #ChineseAI #OpensourceAI

We explore why stories like Pinterest saving $10 million by moving to open models are real, but often misunderstood, and why timing matters more than ideology. Raffi lays out his view of a missing “LAMP stack for AI” and explains why the hardest problem to solve isn’t models or data, but the connective glue that holds AI systems together.

Along the way, he shares how Mozilla is navigating these tradeoffs in practice, why even open-source-first organizations still rely on closed tools during experimentation, and what the browser era taught Mozilla about defaults, user choice, and long-term control.

He also shares a few practical recommendations in this episode that apply even if you’re still experimenting. Listen closely.

This conversation kicks off our Open Source AI series for 2026, focused on real tradeoffs, real economics, and the decisions companies are making right now. Follow on: https://www.turingpost.com/

*Did you like the episode? You know the drill:*

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:* Raffi Krikorian, CTO at Mozilla

LinkedIn: https://www.linkedin.com/in/rkrikorian/

Mozilla AI: https://mozilla.ai/

Mozilla Blog: https://blog.mozilla.org/en/mozilla/mozilla-open-source-ai-strategy/

*Links mentioned:*

Raffi's post here about our OSAI strategy: https://blog.mozilla.org/en/mozilla/mozilla-open-source-ai-strategy/

🌐 #1: Mastering Open Source AI in 2026: Essential Decisions for Builders https://www.turingpost.com/p/opensource1

Mozilla Data Collective: https://data.mozilla.org/

Langchain: https://www.langchain.com/

OpenRouter: https://openrouter.ai/

AI2 (Allen Institute for AI): https://allenai.org/

Flower AI (Federated Learning): https://flower.dev/

Einstein's Dreams by Alan Lightman: https://www.goodreads.com/book/show/14376.Einstein_s_Dreams

📰 The transcript and edited version at https://www.turingpost.com/krikorian

*Chapters:*

0:00 Cold Open — Values vs Economics in Open Source AI

0:28 Intro: Why This Season Focuses on Open Source AI

0:54 When Open Source Becomes a Business Decision

1:44 Pinterest Saved $10M + The Shift From Prototyping to Production

2:42 Mozilla’s “Choice Suite” + The Terraform “Exit Door”

5:21 Mozilla’s Mission: Do for AI What Mozilla Did for the Web

7:09 The “LAMP Stack” for AI + Standards Across the Stack

9:52 Small Models, Specialization, and Model Composability

15:45 Data, Privacy, and “I Own My Context”

18:36 “This Is a Fight Worth Having” + The Signal Analogy

21:42 1–2–3 Steps for Companies to Start (Instrument Choice Early)

24:22 Book Pick: Einstein’s Dreams + Closing

Turing Post is a newsletter about AI's past, present, and future. Ksenia Se explores how intelligent systems are built – and how they're changing how we think, work, and live.

*Follow us →*

Turing Post: https://x.com/TheTuringPost

Ksenia Se: https://www.linkedin.com/in/ksenia-se

https://huggingface.co/Kseniase

#OpenSourceAI #LAMPStackForAI #AIEconomics #MozillaAI #AIInfrastructure #DataProvenance #FederatedLearning #OpenModels

Carina Hong, co-founder and CEO of Axiom Math, argues that to build true reasoning capabilities, we need to move beyond "chatty" models to systems that can verify their own work using formal logic.

In this episode of Inference, we get into:

Why current LLMs are like secretaries (good at retrieval) but bad at de novo mathematics

The three pillars of an AI Mathematician

How AlphaGeometry proved that symbolic logic and neural networks must merge

The difference between AGI and Superintelligence

Why "Theory Building" is harder to benchmark than the International Math Olympiad (IMO)

The scarcity of formal math data (Lean) compared to Python code

We also discuss the bottlenecks: the "chicken and egg" problem of auto-formalization, why Axiom bets on specific superintelligence over general models, and how AI will serve as the algorithmic pillar for the future of hard science.

This is a conversation about the structure of truth, the limits of intuition, and what happens when machines start grading their own homework. Watch it!

Did you like the episode? You know the drill:

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

*Guest:*

Carina Hong, co-founder and CEO of Axiom Math

https://www.axiom.xyz/

https://x.com/CarinaLHong

https://www.linkedin.com/in/carina-hong/

📰 The transcript and edited version at https://www.turingpost.com/carina/

Chapters:

0:53 Why LLMs Struggle with Basic Math

2:42 Building an AI Mathematician: The 3 Pillars (Prover, Knowledge Base, Conjecturer)

5:50 The Role of Human-AI Collaboration

6:34 Can AI Have Intuition? (Conjectures & AlphaGeometry)

10:16 A Hybrid Approach: LLMs + Formal Verification

11:24 Specialist Science Models vs. Generalist Giants

13:33 The Problem with Current AI Benchmarks

16:34 Practical Applications: Enterprise & Formal Verification

21:24 The Main Bottleneck: Data Scarcity

23:49 AGI vs. Superintelligence: The "Plate" Analogy

26:31 Book Recommendations (Math, Law, and Literature)

30:56 How to Use AI for Math Discovery Today

Turing Post is a newsletter about AI's past, present, and future. Ksenia Se explores how intelligent systems are built – and how they're changing how we think, work, and live.

Follow us →

Ksenia and Turing Post:

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

https://huggingface.co/Kseniase

#AI #FutureOfAI #MathAI #FormalVerification #Lean #AxiomMath #Superintelligence #Reasoning

Anneka Gupta, Chief Product Officer at Rubrik, argues we're now living in the world before the crash – where autonomous systems can create their own failures.

In this episode of Inference, we explore:

Why AI agents are "the human problem on steroids"

The three pillars of AI resilience: visibility, governance, and reversibility

How to log everything an agent does (and why that's harder than it sounds)

The mental shift from deterministic code to outcome-driven experimentation

Why most large enterprises are stuck in AI prototyping (70-90% never reach production)

The tension between letting agents act and keeping them safe

What an "undo button" for AGI would actually look like

How AGI will accelerate the cat-and-mouse game between attackers and defenders

We also discuss why teleportation beats all other sci-fi tech, why Asimov's philosophical approach to robots shaped her thinking, and how the fastest path to AI intuition is just... using it every day.

This is a conversation about designing for uncertainty, building guardrails without paralyzing innovation, and what security means when the system can outsmart its own rules.

Did you like the episode? You know the drill:

📌 Subscribe for more conversations with the builders shaping real-world AI.

💬 Leave a comment if this resonated.

👍 Like it if you liked it.

🫶 Thank you for watching and sharing!

Guest: Anneka Gupta, Chief Product Officer at Rubrik https://www.linkedin.com/in/annekagupta/

https://x.com/annekagupta

https://www.rubrik.com/

📰 Want the transcript and edited version?

Subscribe to Turing Post: https://www.turingpost.com/subscribe

Chapters:

Turing Post is a newsletter about AI's past, present, and future. Ksenia Se explores how intelligent systems are built – and how they're changing how we think, work, and live.

Follow us →

Ksenia and Turing Post:

https://x.com/TheTuringPost

https://www.linkedin.com/in/ksenia-se

https://huggingface.co/Kseniase

#AI #AIAgents #Cybersecurity #AIGovernance #EnterpriseAI #AIResilience #Rubrik #FutureOfSecurity