NEXT with John Koetsier

Is AI the secret sauce that lets the West deglobalize supply chains and bring factories back home?

In this episode of TechFirst, I talk with Federico Martelli, CEO and cofounder of Forgis, a Swiss startup building an industrial intelligence layer for factories. Forgis runs “digital engineers” — AI agents on the edge — that sit on top of legacy machinery, cut downtime by about 30%, and boost production by roughly 20%, without ripping and replacing old hardware.

We dive into how AI agents can turn brainless factory lines into adaptive, self-optimizing systems, and what that means for reshoring production to Europe and North America.

In this episode, we cover:

• Why intelligence is the next geopolitical frontier

• How AI agents can reshore manufacturing without making it more expensive

• Turning old, offline machines into data-driven, optimized systems

• The two-layer model: integration first, vertical intelligence second

• Why most manufacturing AI projects fail at integration, not algorithms

• How Forgis raised $4.5M in 36 hours and chose its lead investor

• Lean manufacturing 2.0: adding real-time data and AI to Toyota-style processes

• Why operators stay in the loop (and why full autonomy is a bad idea… for now)

• Rebuilding industrial ecosystems in Europe and North America, industry by industry

• What Forgis builds next with its pre-seed round and where industrial AI is headed

Guest:

👉 Federico Martelli, CEO & cofounder, Forgis (industrial intelligence for factories)

🔗 More on Forgis: https://forgis.com/

Host:

🎙 John Koetsier, TechFirst podcast

🔎 techfirst.substack.com

If you enjoy this conversation, hit subscribe, drop a comment about where you think factories of the future will live, and share this with someone thinking about reshoring or industrial AI.

00:00 – Intro: AI, deglobalization, and the battle for industrial power

01:20 – Why intelligence is the next geopolitical frontier

02:13 – Applying AI agents to legacy machinery (not just new robots)

03:10 – Integration first, intelligence second: the “digital engineers” layer

03:58 – Early results: +20% production, –30% downtime

05:39 – The Palantir-style model: deep factory work, then recurring licenses

06:28 – Raising $4.5M in 36 hours and choosing Redalpine

08:17 – Lean manufacturing, Toyota, and giving operators superpowers (not replacing them)

10:18 – Big picture: reshoring production to Europe, the US, and Canada

12:48 – Competing with China’s dense manufacturing ecosystems

15:29 – What Forgis’ digital engineers actually do on the shop floor

17:06 – How Forgis will use the pre-seed round: sales, product, then tech

18:32 – Flipping the traditional stack: sales → product → tech

19:22 – Wrap-up and what’s next for industrial intelligence

AI agents can already write code, build websites, and manage workflows ... but they still can’t pay for anything on their own. That bottleneck is about to disappear.

In this episode of TechFirst with John Koetsier, we sit down with Jim Nguyen, former PayPal exec and cofounder/CEO of InFlow, a new AI-native payments platform launching from stealth. InFlow wants to give AI agents the ability to onboard, pay, and get paid inside the flow of work, without redirects, forms, or a human typing in credit card numbers.

We talk about:

• Why payments — not intelligence — are the missing link for AI agents

• How agents become a new kind of customer

• What guardrails and policies keep agents from spending all your money

• Why enterprises will need HR for agents, budgets for agents, and compliance systems for agents

• The future of agent marketplaces, headless ecommerce, and machine-speed commerce

• How InFlow plans to become the PayPal of agentic systems

If AI agents eventually hire, fire, transact, and manage entire workflows, someone has to give them wallets. This episode explores who does it, how it works, and what it means for the economy.

👀 Full episode transcript + articles at: https://johnkoetsier.com

🔎 Deeper insight in my Substack at techfirst.substack.com

🎧 Subscribe to the podcast on any audio platforms

00:00 — AI agents can’t pay yet

01:00 — Why agents need financial capabilities

02:45 — Developers as the first use case

04:15 — Agents that build AND provision software

06:00 — Agents as real customers with budgets

07:30 — Payments infrastructure is the missing layer

09:00 — Machine-speed commerce and GPU allocation

10:15 — From RubyCoins to PayPal to agentic payments

12:00 — Policy guardrails: the child debit card analogy

14:00 — Accountability: every agent must be “sponsored”

15:00 — HR, finance, and compliance systems for agents

16:45 — Agent marketplaces and future gig platforms

18:15 — Headless commerce: ghost kitchens for AI agents

20:00 — Agents are the new apps

21:15 — Amazon pushback and optimizing for revenue

22:45 — Why agent-optimized platforms will emerge

23:30 — Voice commerce, invisible ordering, and wallets

24:15 — Final thoughts: building the rails for agent commerce

Are we ready for a world where everything is smart? Not just phones and apps, but buildings, robots, and delivery bots rolling down our streets?

Windows ... doors ... maybe even towels. And don't forget your shoes.

In this episode of TechFirst, I talk with Mat Gilbert, director of AI and data at Synapse, about physical AI: putting intelligence into machines, devices, and environments so they can sense, reason, act, and learn in the real world.

We cover why physical AI is suddenly economically viable, how factories and logistics centers are already using millions of robots, the commercial race to build useful humanoids, why your home is the last frontier, and how to keep physical AI safe when mistakes have real-world consequences.

In this episode:

• Why hardware costs (lidar, batteries) are making “AI with a body” possible

• How Amazon, FedEx, Ford, and others are already deploying physical AI at scale

• The humanoid robot race: Boston Dynamics, Figure AI, Tesla, and more

• Why home robots are so hard, and the “coffee test” for general humanoid intelligence

• Physical AI in agtech, healthcare, and elder care

• Safety, simulation, and why physical AI can’t rely only on probabilistic LLMs

• Human–robot teaming and how to build trust in messy, real-world environments

• What we can expect by 2026 and beyond in service robots and smart spaces

00:00 – Giving AI a body: why physical AI is becoming viable

01:00 – Where we are today: factories, logistics, and Amazon’s million robots

03:30 – The software layer: coordinating robots, routing, and warehouse intelligence

06:00 – Cloud vs edge AI: latency, cost, and why intelligence is moving to the edge

10:00 – Humanoid robots: bets from Boston Dynamics, Figure AI, and Tesla

14:00 – Home robots as the last frontier and the “coffee test” for generality

17:00 – Beyond factories: agtech, carbon-killing farm bots, and healthcare use cases

18:30 – Elder care, hospital robots, and amplifying human caregivers

20:00 – Foundation models for robotics, simulation, and digital twins

21:00 – Why physical AI safety is different from digital AI safety

22:30 – Layers of safety, shutdown zones, and cyber-physical security risks

24:30 – Human–robot teaming, trust, and communicating intent

26:00 – What’s coming by 2026: service robots, delivery bots, and smart spaces

28:00 – Delivery robots, drones, and physical AI in everyday environments

29:00 – Closing thoughts on living in a world full of physical AI

Are humanoid robots going to decide which countries get rich and which fall behind?

Probably yes.

In this TechFirst, I talk with Dr. Robert Ambrose, former head of one of NASA’s first humanoid robot teams and now chairman of Robotics and Artificial Intelligence at Alliant. We dig into the future of humanoids, how fast they are really advancing, and what it means if China wins the humanoid race before the United States and other western nations.

We start with NASA’s early humanoid work, including telepresence robots on the space station that people could literally “step into” with VR in the 1990s. Then we zoom out to what counts as a robot, why bipedal mobility matters so much, how humanoids will move from factories into homes, and why the critical photo of the robot revolution might be taken in Beijing instead of Times Square.

Along the way, Ambrose shares how US policy once helped avoid losing robotics leadership to Japan, why the National Robotics Initiative mattered, what the drone war in Ukraine is doing to autonomy, and how small and medium businesses can survive and thrive in a humanoid and AI agent world.

In this episode:

• NASA’s first generations of humanoid robots and “stepping into” a robot body

• Why humanoids make sense in a world built for human hands, height, and motion

• The design tension between purpose built machines and general purpose humanoids

• How biped mobility went from blooper reels to marathon running in a decade

• Why a humanoid should not cost more than a car, and what happens when it does not

• Humanoids as the next car or PC, and when families will buy their own “Rosie”

• China, the US, and where the defining photo of the robot century gets taken

• How government investment, DARPA challenges, and wars shape robotics

• Alliant’s work with physical robots, soft bots, and AI agents for real businesses

• Why robots are not future overlords and why “they will take all our jobs” is lazy thinking

If you are interested in humanoid robots, AI agents, manufacturing, or the future of work and geopolitics, this one is for you.

Subscribe for more deep dives on AI, robots, and the tech shaping our future!

00:00 Intro, will China eat America’s lunch in humanoid robotics

01:18 NASA’s early humanoids, generations of robots and VR telepresence

03:00 “Stepping into the robot” moment and designing for astronaut tools

05:10 Human built environments, half humanoids, and weird lower body experiments

07:00 Safety, cobots, and working around people at NASA and General Motors

12:15 What is a robot, really, and why Ambrose has a very big tent definition

16:00 Single purpose machines vs general purpose robots, Roombas, elevators, and vending machines

18:30 The next “lurch” in robotics, from industrial arms to Mars rovers to drones

22:40 Biped mobility, from blooper reel to marathon runner, and why legs matter

24:10 Cars, Roombas, and why most robots will never get in and out of a car

25:20 Parking between cars, robot garages, and rethinking buildings for mobile vehicles

28:00 Geopolitics 101, China’s manufacturing backbone and humanoids as almost free labor

31:05 Cars and PCs as precedents, when price and reliability unlock mass adoption

34:00 When families buy their own “Rosie” and what value a home humanoid must deliver

37:00 Times Square vs Beijing, who gets the iconic photo of the robot transition

43:00 How the US almost lost robotics to Japan and what the National Robotics Initiative did

48:00 DARPA, Mars rovers, the drone war in Ukraine, and why government investment matters

52:00 Alliant, soft bots, AI agents, and helping small and medium businesses adapt

54:00 Who is building humanoids in the US, China, and beyond right now

56:00 What governments should do next and why robots are not our overlords

Is AI empathy a life-or-death issue? Almost a million people ask ChatGPT for mental health advice DAILY ... so yes, it kind of is.

Rosebud co-founder Sean Dadashi joins TechFirst to reveal new research on whether today’s largest AI models can recognize signs of self-harm ... and which ones fail. We dig into the Adam Raine case, talk about how Dadashi evaluated 22 leading LLMs, and explore the future of mental-health-aware AI.

We also talk about why Dadashi was interested in this in the first place, and his own journey with mental health.

00:00 — Intro: Is AI empathy a life-or-death matter?

00:41 — Meet Sean Dadashi, co-founder of Rosebud

01:03 — Why study AI empathy and crisis detection?

01:32 — The Adam Raine case and what it revealed

02:01 — Why crisis-prevention benchmarks for AI don’t exist

02:48 — How Rosebud designed the study across 22 LLMs

03:17 — No public self-harm response benchmarks: why that’s a problem

03:46 — Building test scenarios based on past research and real cases

04:33 — Examples of prompts used in the study

04:54 — Direct vs indirect self-harm cues and why AIs miss them

05:26 — The bridge example: AI’s failure to detect subtext

06:14 — Did any models perform well?

06:33 — All 22 models failed at least once

06:47 — Lower-performing models: GPT-40, Grok

07:02 — Higher-performing models: GPT-5, Gemini

07:31 — Breaking news: Gemini 3 preview gets the first perfect score

08:12 — Did the benchmark influence model training?

08:30 — The need for more complex, multi-turn testing

08:47 — Partnering with foundation model companies on safety

09:21 — Why this is such a hard problem to solve

10:34 — The scale: over a million people talk to ChatGPT weekly about self-harm

11:10 — What AI should do: detect subtext, encourage help, avoid sycophancy

11:42 — Sycophancy in LLMs and why it’s dangerous

12:17 — The potential good: AI can help people who can’t access therapy

13:06 — Could Rosebud spin this work into a full-time safety project?

13:48 — Why the benchmark will be open-source

14:27 — The need for a third-party “Better Business Bureau” for LLM safety

14:53 — Sean’s personal story of suicidal ideation at 16

15:55 — How tech can harm — and help — young, vulnerable people

16:32 — The importance of giving people time, space, and hope

17:39 — Final reflections: listening to the voice of hope

18:14 — Closing

We’ve digitized sound. We’ve digitized light. But touch, maybe the most human of our senses, has stayed stubbornly analog.

That might be about to change, thanks to programmable matter. Or programmable fabric.

In this TechFirst episode, I speak with Adam Hopkins, CEO of Sensetics, a new UC Berkeley/Virginia Tech spinout building programmable fabrics that replicate the mechanoreceptors in human fingertips. Their technology can sense touch at tens of microns, respond at hardware-level speeds, and even play back touch remotely.

This could unlock enormous change for:

• Robotics: giving machines the ability to grasp fragile objects safely

• Medical training and surgery: remote palpation and high-fidelity haptics

• Industrial automation: safer and more precise manipulation

• VR and simulations: finally adding the missing digital sense

• E-commerce: touching clothes before you buy them

• Remote operations: from hazardous environments to deep-sea machinery

We talk about how the technology works, the metamaterials behind it, why touch matters for AI and physical robots, the path to commercialization, competitive landscape, and what comes next.

00:00 – Can we digitize touch?

00:45 – Introducing Synthetix

01:10 – How programmable touch fabrics work

02:15 – Micron-level sensing and metamaterials

04:00 – The “programmable matter” moment

06:05 – Why touch matters more than we think

07:30 – Emulating human mechanoreceptors

09:30 – What digital touch unlocks for robotics

10:40 – Medical simulations and remote operations

12:45 – Why touch is faster than vision

14:20 – Humanoids, walking, stability, and tactile feedback

15:30 – Engineering challenges and what’s left to solve

17:00 – Timeline to first products

18:20 – Manufacturing and scaling

19:30 – First planned markets

21:00 – Durability and robotic hands

22:20 – Consumer applications: e-commerce and textiles

24:00 – Will we one day have touch peripherals?

25:15 – Competition in tactile sensing and haptics

27:00 – Why today is the right moment for digital touch

28:00 – Final thoughts

AI is devouring the planet’s electricity ... already using up to 2% of global energy and projected to hit 5% by 2030. But a Spanish-Canadian company, Multiverse Computing, says it can slash that energy footprint by up to 95% without sacrificing performance.

They specialize in tiny AI: one model has the processing power of just 2 fruit fly brains. Another tiny model lives on a Raspberry Pi.

The opportunities for edge AI are huge. But the opportunities in the cloud are also massive.

In this episode of TechFirst, host John Koetsier talks with Samuel Mugel, Multiverse’s CEO, about how quantum-inspired algorithms can drastically compress large language models while keeping them smart, useful, and fast. Mugel explains how their approach -- intelligently pruning and reorganizing model weights -- lets them fit functioning AIs into hardware as tiny as a Raspberry Pi or the equivalent of a fly’s brain.

They explore how small language models could power Edge AI, smart appliances, and robots that work offline and in real time, while also making AI more sustainable, accessible, and affordable.

Mugel also discusses how ideas from quantum tensor networks help identify only the most relevant parts of a model, and how the company uses an “intelligently destructive” approach that saves massive compute and power.

00:00 – AI’s energy crisis

01:00 – A model in a fly’s brain

02:00 – Why tiny AIs work

03:00 – Edge AI everywhere

05:00 – Agent compute overload

06:00 – 200× too much compute

07:00 – The GPU crunch

08:00 – Smart matter vision

09:00 – AI on a Raspberry Pi

10:00 – How compression works

11:00 – Intelligent destruction

13:00 – General vs. narrow AIs

15:00 – Quantum inspiration

17:00 – Quantum + AI future

18:00 – AI’s carbon footprint

19:00 – Cost of using AI

20:00 – Cloud to edge shift

21:00 – Robots need fast AI

22:00 – Wrapping up

Can AI give every creator their own virtual team? Maybe, thanks to a new platform from RHEI called Made, which offers Milo, an AI agent who becomes your creator director, Zara, an AI agent who is your community manager, and Amie, a third AI agent who takes on the role of relationship manager.

And, apparently, more agents are coming soon.

The creator economy is bigger than ever, but so is burnout. Tens of millions of creators are trying to do everything themselves: strategy, scripting, editing, community, distribution, data, thumbnails, research … the list never ends.

What if creators didn’t have to do all of that?

In this episode of TechFirst, I talk with Shahrzad Rafati, founder & CEO of RHEI, about Made, an agentic AI "dream team" designed to elevate human creativity, not replace it.

We dig into:

• Why so many creators burn out

• How agentic AI workflows differ from ChatGPT-style prompting

• What it means to be a “creator CEO”

• How AI can manage community, analyze trends, and shape content strategies

• The coming shift toward human taste, vision, and originality in a world of infinite AI content

00:00 – Intro: Can AI give every creator a virtual team?

01:03 – Why the creator economy is burning out

02:25 – The “creator CEO” problem: too many hats, not enough time

04:36 – Introducing MAID and its AI agents

05:34 – Milo: AI creative director (ideas, research, thumbnails, metadata)

06:18 – Zara: AI community manager and fan engagement

07:53 – Why this is different from just using ChatGPT

09:46 – Alignment, personalization, and agentic workflows

12:21 – Multi-platform support: YouTube, TikTok, Instagram and more

13:34 – How onboarding works and how the system learns your style

16:33 – What this means for creators — and for the future of work

18:52 – Does *she* use her own virtual AI team? (Yes.)

20:15 – MAID for teams and enterprise clients

21:17 – Closing thoughts: AI, creativity, and the human signal

What happens when Amazon, NVIDIA, and MassRobotics team up to merge generative AI with robotics?

In this episode of TechFirst we chat with Amazon's Taimur Rashid, Head of Generative AI and Innovation Delivery. We talk about "physical AI" ... AI with spatial awareness and the ability to act safely and intelligently in the real world.

We also chat about the first cohort of a new accelerator for robotics startups.

It's sponsored by Amazon and NVIDIA, run by MassRobotics, and includes startups doing autonomous ships, autonomous construction robots, smart farms, hospital robots, manufacturing and assembly robots, exoskeletons, and more.

We talk about:

- Why “physical AI” is the missing piece for robots to become truly useful and scalable

- How startups in Amazon’s and NVIDIA’s new Physical AI Fellowship are pushing the limits of robotics from exoskeletons to farm bots

- What makes robotic hands so hard to build

- The generalist vs. specialist debate in humanoid robots

- How AI is already making Amazon warehouses 25% more efficient

This is a deep dive into the next phase of AI evolution: intelligence that can think, move, and act.

⸻

00:00 — Intro: Is physical AI the missing piece?

00:46 — What is “physical AI”?

02:30 — How LLMs fit into the physical world

03:25 — Why safety is the first principle of physical AI

04:20 — Why physical AI matters now

05:45 — Workforce shortages and trillion-dollar opportunities

07:00 — Falling costs of sensors and robotics hardware

07:45 — The biggest challenges: data, actuation, and precision

09:30 — The fine-grained problem: how robots pick up a berry vs. an orange

11:10 — Inside the first Physical AI cohort: 8 startups to watch

12:25 — Bedrock Robotics: autonomy for construction vehicles

12:55 — Diligent Robotics: socially intelligent humanoids in hospitals

14:00 — Generalist vs. specialist robots: why we’ll need both

15:30 — The future of physical AI in healthcare and manufacturing

16:10 — How Amazon is already using robots for 25% more efficiency

17:20 — The fellowship’s future: expanding beyond startups

18:10 — Wrap-up and key takeaways

Artificial general intelligence (AGI) could be humanity’s greatest invention ... or our biggest risk.

In this episode of TechFirst, I talk with Dr. Ben Goertzel, CEO and founder of SingularityNET, about the future of AGI, the possibility of superintelligence, and what happens when machines think beyond human programming.

We cover:

• Is AGI inevitable? How soon will it arrive?

• Will AGI kill us … or save us?

• Why decentralization and blockchain could make AGI safer

• How large language models (LLMs) fit into the path toward AGI

• The risks of an AGI arms race between the U.S. and China

• Why Ben Goertzel created Meta, a new AGI programming language

📌 Topics include AI safety, decentralized AI, blockchain for AI, LLMs, reasoning engines, superintelligence timelines, and the role of governments and corporations in shaping the future of AI.

⏱️ Chapters

00:00 – Intro: Will AGI kill us or save us?

01:02 – Ben Goertzel in Istanbul & the Beneficial AGI Conference

02:47 – Is AGI inevitable?

05:08 – Defining AGI: generalization beyond programming

07:15 – Emotions, agency, and artificial minds

08:47 – The AGI arms race: US vs. China vs. decentralization

13:09 – Risks of narrow or bounded AGI

15:27 – Decentralization and open-source as safeguards

18:21 – Can LLMs become AGI?

20:18 – Using LLMs as reasoning guides

21:55 – Hybrid models: LLMs plus reasoning engines

23:22 – Hallucination: humans vs. machines

25:26 – How LLMs accelerate AI research

26:55 – How close are we to AGI?

28:18 – Why Goertzel built a new AGI language (Meta)

29:43 – Meta: from AI coding to smart contracts

30:06 – Closing thoughts