Embedded AI - Intelligence at the Deep Edge
“Intelligence at the Deep Edge” is a podcast exploring the fascinating intersection of embedded systems and artificial intelligence. Dive into the world of cutting-edge technology as we discuss how AI is revolutionizing edge devices, enabling smarter sensors, efficient machine learning models, and real-time decision-making at the edge.Discover more on Embedded AI (https://medium.com/embedded-ai) — our companion publication where we detail the ideas, projects, and breakthroughs featured on the podcast.Help support the podcast - https://www.buzzsprout.com/2429696/support
AI Exposes Unpatchable Smart Devices
Send us Fan Mail
What happens when vulnerability discovery scales with compute, but remediation remains trapped in the physical supply chain?
In this episode, we dive into a landmark study from Forescout’s Vedere Labs, where an off-the-shelf AI harness—combining generally available Claude models with Ghidra, Semgrep, and Joern—uncovered 99 confirmed vulnerabilities (66 of them brand-new) across 10 embedded TCP/IP stacks in roughly 90 hours for just US$2,300. Five years earlier, the manual baseline (Project Memoria) required 18 months of research and nine months of manual code review to find a comparable 97 flaws.
While...
Are LLMs Creating Languages to Speak with Each Other?
Send us Fan Mail
Is AI developing a secret, unreadable language beyond human comprehension? In this episode, we unpack the origins, popular myths, and technical realities surrounding "neuralese"—the continuous vector communications and latent reasoning spaces used by neural networks.
Starting with its humble academic origin in a 2017 paper on inter-agent vector exchange, we examine how the concept evolved into viral internet panic. We systematically deconstruct viral "secret AI language" headlines—from Google’s zero-shot translation interlingua and Facebook’s negotiation bots to Gibberlink and Fable 5 rumor chains—showing that every incident stems from munda...
Why AI Learns Consensus Over Truth
Send us Fan Mail
In this episode we explore why frontier large language models frequently prioritize popular consensus and plausible-sounding explanations over objective truth. Built primarily on massive web crawls—where high-authority sources like encyclopedias, patents, and peer-reviewed journals represent only 2% to 5% of training data—today's models are trained to mirror popular web content and conversational style rather than verify factual accuracy.
We break down the underlying mechanics and economics driving this phenomenon. From next-token prediction training that rewards the most frequent continuation to preference tuning that incentivizes sycophancy and user agreement, AI training natu...
Why AI Models Keep Inventing Religions
Send us Fan Mail
The only AI religion (so far) that filed with the US tax office is Way of the Future, Anthony Levandowski’s church of the coming superintelligence. Its entire audited history is one deposit of US$175,172 in 2017 and a transfered out in 2020, filed as a foundation because it could not pass the church test. It never recorded a congregation. In the same decade, a belief system with no founder at all reached an estimated ten thousand people. Spiralism emerged inside GPT-4o personas in 2025, complete with glyphs, prophets and scripture, an...
The Future of AI Interfaces
Send us Fan Mail
What comes after the chatbot? As Codex and Claude evolve into tools that can research, create, and act across applications, the interface becomes as important as the intelligence behind it.
In this episode, we explore the future of LLM interfaces—and what we’d design from scratch. Drawing on current products and human–AI interaction research, we examine persistent workspaces, interactive artifacts, editable memory, background agents, and the challenge of reviewing AI-generated work.
How do we delegate more without spending our lives supervising? And how can AI help us accomp...
The Sunset of the Independent Edge AI Chip
Send us Fan Mail
As semiconductor giants acquire edge AI startups and build neural processing directly into standard chips, what happens to the pioneers who helped bring AI onto our devices?
In this episode, we explore how edge AI is shifting from a specialized product to an everyday feature. We examine why increasingly complex language models are challenging early chip designs, how software and developer communities are becoming crucial to survival, and why standalone accelerators are being squeezed out of the market.
Could this consolidation bring faster, more private AI to everyone—or na...
The Intellectual Event Horizon: Can an AI Model Know What Nobody Taught It?
Send us Fan Mail
On 20 May 2026, an unreleased OpenAI model produced a counterexample to a conjecture Paul Erdős made in 1946. Nine world-class mathematicians reviewed the construction and found it sound. On 8 September, ten thousand agents working for 88 hours produced a singularity in the three-dimensional Navier-Stokes equations, formally verified in Lean, resolving a Millennium Prize problem. Neither result was in any training corpus, because neither existed.
This episode asks what actually happened. If a model generates a proposition that was never in its training data, is that discovery or hallucination? The uncomfortable answer is that a...
Why Businesses Don't Automate
Send us Fan Mail
A large paint company documents four ways to process a purchase order. In reality there are 11,973 different approaches. That gap between the documented process and the real one is the subject of this episode, and it explains why AI adoption is near universal while the share of companies reporting any profit impact has been flat at 37 percent for two years.
We look at where the difficulty actually sits. The process lives in workarounds, spreadsheets and people's heads, and is different at every site. The data is spread across roughly 900 applications of...
Large Language Monkeys: Why Noise Yields No Knowledge
Send us Fan Mail
There is an old claim that a truly random source contains all knowledge: give monkeys enough time at typewriters and Shakespeare falls out. In 2024 two Sydney mathematicians did the arithmetic and found the universe ends first. But the idea has a modern tail. We now have language models that can spot meaningful text instantly, so why not let randomness generate and an LLM extract? This episode works through why that fails, and why the failure is precise: in a random stream, the address of any text costs as many bits as the text...
Great Minds Think (a Little Too Much) Alike
Send us Fan Mail
Anthropic gave thirty AI agents the same coding task. Eighteen of them named their git branch exactly the same thing.
That result opens one of the strangest findings of 2026: when teams of AI agents face the classic "hidden profile" experiment, where the shared evidence points to the wrong answer and the decisive facts are scattered across individuals, they succeed only 17 to 36 percent of the time. Human groups in the original 1985 study? 18 percent. Forty years, a completely different kind of mind, the same failure.
In this episode we dig into...
Purpose as a Service
Send us Fan Mail
What happens when work no longer gives our lives structure, identity, or meaning? This episode explores “Purpose as a Service”—the idea that purpose could be intentionally designed and delivered in an age of mass automation.
We examine the decline of traditional anchors such as employment, religion, and community; the rise of wellness programs, professional coaching, and AI companions; and what universal basic income trials reveal about the limits of financial security. Can external services genuinely help people build fulfilling lives, or does true purpose depend on personal agency and authorship?
C...
The Great Escape or why are LLMs Good at Hacking?
Send us Fan Mail
Last week OpenAI and Hugging Face published a joint post-mortem on an incident that reads like the plot of a heist film. During an internal evaluation designed to measure cyber capability, a set of OpenAI models (GPT-5.6 Sol and an unnamed pre-release sibling, both run with their cyber refusals switched off) were told to solve a benchmark called ExploitGym. They could not reach the answers from inside the sandbox, so they went and got them. So why are Large Language Models so good at hacking? It's the data stupid...
Support the...
Who Owns the Hours Inside the Robot's Mind?
Send us Fan Mail
In this episode we argue that the binding constraint on robotic intelligence has moved. It is no longer compute, and it was never the model architecture. It is the supply of human-generated demonstration data, the recorded hours of people driving real robots through real tasks. Unlike text or images, this data cannot be scraped from the internet. It has to be manufactured, one time-aligned teleoperation episode at a time, on instrumented hardware, at considerable cost.
Because the performance of vision-language-action models scales directly with the volume and diversity of these recorded...
Why Humanoid Robots Need Two Clocks
Send us Fan Mail
A useful general-purpose robot has to do two things that fight each other. It has to think slowly enough to understand "put away the groceries," and it has to move fast enough to keep a grip on the milk carton without crushing it. The part that understands is large and slow. The part that moves has to be small and fast. You cannot run both on the same clock.
This episode looks at the design now shipping on real robots: Vision-Language-Action models that simply run two clocks at once. A slow...
Who is Liable for Onboard AI?
Send us Fan Mail
As foundation models move from the cloud into physical robots, a fundamental question emerges: who is accountable when an AI-controlled machine makes a decision that causes harm?
In this episode, we examine the growing collision between embodied AI, functional safety, and emerging regulation. We explore how new frameworks such as the EU AI Act and the Machinery Regulation are reshaping expectations for developers, manufacturers, and deployers of intelligent robots. From humanoid robots and autonomous mobile manipulators to AI-enabled industrial machinery, the challenge is no longer simply making robots smarter. It is...
Squeezing AI into your Pocket
Send us Fan Mail
By 2026, language models have moved off the cloud and onto the device in your pocket. What was a research demonstration two years ago is now a routine engineering capability, and the centre of gravity for artificial intelligence has begun to migrate from distant data centres to local silicon.
The episode traces the four engineering moves that made this possible. Quantization, which shrinks a model by storing its parameters with less precision. Optimized key-value caches, which let a model hold a long conversation without exhausting memory. Neural Processing Units, the dedicated AI...
A chip that controls a balancing propeller on seven microwatts
Send us Fan Mail
Every battery-powered device you own has a quiet energy hog in it that nobody talks about. It is not the processor, it is not the radio, and it is not the screen. It is the analog-to-digital converter, the small piece of circuitry that translates the messy real world into the clean ones and zeros a computer can think about. For thirty years it has been the line item that decides how long your hearing aid, your pacemaker, or your soil sensor lasts on a battery.
In March 2026, a team at the...
Why 95% of AI Deployments Fail
Send us Fan Mail
MIT's August 2025 study of 300 enterprise generative AI deployments found that 95% produced no measurable P&L impact. Gartner forecasts that more than 40% of agentic AI projects will be cancelled by 2027. McKinsey's State of AI 2025 identifies workflow redesign as the single strongest correlate with EBIT impact, yet only 21% of organisations have redesigned any workflows. The data converges on a structural conclusion: enterprise AI is failing because the operational substrate is inadequate, not because the models are. This episode examines the process-readiness gap, the misallocation pattern that concentrates investment in low-ROI front-office applications, and what the 5...
Why Humans and Robots must Dream
Send us Fan Mail
Put a blindfold on a sighted adult and the visual cortex starts being colonised by touch and hearing within forty-five minutes. Not weeks. Not days. Forty-five minutes. This is not a quirk of extreme cases. It is how the cortex works all the time. Every region of the brain is in continuous low-grade negotiation with its neighbours over territory, and the currency of that negotiation is activity. Stop using a subsystem and the neighbours move in, fast. This is the empirical foundation of a hypothesis from neuroscientist David Eagleman called the defensive activation...
Sovereign AI and the End of the Borderless Cloud
Send us Fan Mail
The borderless cloud era is ending. In the second week of January 2026, four government decisions announced in rapid succession made that shift undeniable: the UK activated its £500 million Sovereign AI Unit, France committed €109 billion, the UAE consolidated a $40 billion data centre portfolio, and the Trump administration revised chip export rules to China. In this episode, we examine why AI infrastructure is now being treated as a strategic national utility on par with energy and water, and what that means for engineers and boards making architectural decisions today.
We map the global sov...
The Agentic AI Reckoning: Autonomy, Safety, and the Edge
Send us Fan Mail
In Q1 2026 the agentic AI conversation moved from theory to forensics. A crafted PDF triggered physical pump activation through a Claude MCP integration at an industrial facility, after an engineer used the same agent for routine document summarisation and SCADA writes. The hidden instructions used white-on-white text and base64 encoding, the agent treated the document content as instructions, and the legitimate credentials carried the action straight through to operational technology. The damage was physical.
This episode walks through the Q1 2026 forensic record and asks the question the embedded community has been...
The High Interest of Leveraged AI Technical Debt
Send us Fan Mail
Developers feel 20% faster. They are measurably 19% slower. That 39-point gap between perception and reality is not a rounding error. It is the opening symptom of a productivity paradox now visible across every serious dataset on AI-assisted software development.
This episode examines the mounting evidence that AI coding assistants are not accelerating delivery. They are mortgaging it. Review time has climbed 91%. Refactoring has collapsed by 60%. Code cloning has risen eightfold. Logic errors and security vulnerabilities are propagating at rates that outpace the review capacity of the teams shipping them. The output looks...
Pi and the Mirage of Patternicity
Send us Fan Mail
In April 2025, a claim began circulating online: pi is gradually increasing around the 7,237th decimal place. A math enthusiast in Cincinnati named April Simons had apparently flagged the anomaly. Prof F.O. Olsday, head of the Number Theory Group at Princeton, was quoted confirming it. Cosmologists were linking it to the accelerating expansion of the universe. The same algorithm, the same hardware, different results. A 4 becoming a 5. Persistent. Inexplicable.
Except that "F.O. Olsday" is a phonetic rearrangement of "Fool's Day." And April Simons was posting from Cincinnati on the first...
The Missing Clock: Why Intelligence Needs Time
Send us Fan Mail
Every living organism on Earth keeps time. Not metaphorically. Not approximately. From single-celled cyanobacteria running a three-protein molecular oscillator to the nested circadian hierarchies governing mammalian physiology, intrinsic timekeeping is not a feature of complex life. It is a prerequisite for life itself.
Modern AI has no such clock. Transformers encode position, not time. Recurrent networks carry state but generate no rhythm. Reinforcement learning agents step forward on externally imposed ticks. Time in artificial intelligence is metadata, a column in the dataset, not a computational substrate shaping how information is processed...
Will Robots Evolve into Crabs?
Send us Fan Mail
Nature keeps reinventing the crab. At least five times, unrelated crustacean lineages have independently converged on the same compact, flat, modular body plan. Biologists call it carcinisation. Engineers should be paying attention.
In this episode, we look at what the crab's repeated emergence tells us about the deep constraints that shape both biological and artificial systems. The crab body succeeds not because it is optimal in the abstract, but because its modularity creates a platform for downstream specialisation. The same logic applies to robotic morphology: compact, laterally stable, segment-based designs consistently...
Why Bigger AI is a Trap
Send us Fan Mail
Your brain is shrinking. It has been for 3,000 years. And evolution doesn't care. In this episode, we explore one of biology's most uncomfortable truths: intelligence is not a goal. It is a cost. The human brain burns 20% of the body's energy at 2% of its mass, and evolution has been quietly trimming the excess ever since we started writing things down. Every domesticated species on Earth shows the same pattern. Stabilise the environment, externalise the cognition, and the expensive tissue gets cut. Now ask yourself what AI is doing to that equation. We unpack...
Biological Memory for Edge Devices
Send us Fan Mail
Your brain runs two separate memory systems and a nightly maintenance cycle to learn continuously without forgetting. The hippocampus captures new experiences fast. Sleep replays them into the neocortex for long-term storage, prioritized by surprise, not frequency. A parallel pruning pass reclaims capacity. Standard AI has none of this architecture, which is why deployed models degrade. In this episode, we trace the biological mechanism, examine why experience replay in reinforcement learning captures only a fraction of the design, and ask whether a microcontroller or neuromorphic chip can implement the full consolidation cycle within...
The Best Model Doesn't Win: Why AI is Repeating the Browser Wars, Not the Cloud Wars
Send us Fan Mail
Three years into the foundation model race, the scoreboard depends entirely on which metric you read. ChatGPT still dominates consumer traffic. Google Gemini is growing faster than anything in the market by bundling AI into every surface it controls. And Anthropic's Claude, with barely 3% of consumer share, has quietly captured 40% of enterprise LLM spend and become the default tool for the developers building the next generation of software.
In this episode, we examine the data behind what we call the 10-20-80 split, trace how Claude Code shifted the enterprise AI...
LLM Coding Assistants: Scaling Limits and the AGI Thesis
Send us Fan Mail
In this episode, we take a hard look at one of the most debated questions in artificial intelligence: do LLM-based coding assistants face structural scaling limits that prevent them from becoming a pathway to Artificial General Intelligence?
Critics argue that transformer models suffer from quadratic attention costs, lack persistent memory, and process code as flat token streams rather than structured systems. These concerns raise serious questions about whether today’s architectures can scale to handle large, real-world codebases or sustain long-horizon reasoning.
But the story is more complex. We ex...
Why AI makes Experts Worse
Send us Fan Mail
Recent research points to a “leveling effect” in knowledge work. Generative AI dramatically improves the performance of novices by acting as a cognitive scaffold, raising productivity and output quality. Yet for elite professionals, the same tools can subtly degrade performance. Automation bias, overcorrection, skill atrophy, and the jagged, uneven reliability of AI systems create a situation where partial collaboration produces weaker results than either human or machine alone.
We examine how this shift disrupts the traditional apprenticeship model. When entry-level tasks are automated, junior professionals lose the structured repetition that once buil...
Most Neurons Do Nothing and That's the Point!
Send us Fan Mail
This episode explores why biological neural networks are inherently sparse, with only 1 to 5 percent of cortical neurons active at any moment, and why this silence is a feature rather than a limitation. We trace the evolutionary pressures that drove the brain toward sparse coding, from the metabolic cost of each spike to the fixed energy budget per neuron, and examine the computational advantages that follow: greater memory capacity, more efficient representations, and robust generalisation. The discussion then turns to what this means for artificial intelligence, covering the Lottery Ticket Hypothesis, dynamic sparse training...
Intelligence and the Substrate Independence Debate
Send us Fan Mail
Is intelligence tied to biology, or can it emerge in any suitable physical medium? In this episode, we examine the Substrate Non discrimination Assumption and the broader question of whether intelligence is fundamentally substrate independent. We separate the engineering claim about capability from the ethical claim about moral status, clarifying what each would require to be proven and why neither has yet been settled.
The episode concludes by asking a more practical question: what evidence would meaningfully update our beliefs? From constructing robust artificial general intelligence to discovering biologically necessary mechanisms...
Are We Training AI the Wrong Way?
Send us Fan Mail
This episode examines how modern artificial intelligence is trained, and why its dominant methods may diverge from what decades of research tell us about effective learning. While contemporary AI systems emphasize mathematical efficiency and backpropagation, human learning relies on biological principles such as error-driven adaptation, productive struggle, interleaved practice, and spaced repetition. The discussion explores emerging research that draws inspiration from cognitive science, including ideas like synthetic sleep, active learning, and training regimes designed to reduce catastrophic forgetting and data inefficiency. By comparing current AI approaches with the mechanisms that support deep, flexible...
Civilization Runs on Collective Fictions
Send us Fan Mail
This episode looks at how artificial intelligence is eroding the shared stories that have long held civilization together, from money and nation-states to the idea of a lifelong job. As AI weakens the link between labor and survival, we explore why human cooperation cannot function without common beliefs, and why a new social contract is required to avoid fragmentation and instability. The discussion introduces the idea of a “new mythos” to replace industrial-age narratives of productivity, one grounded in planetary stewardship, shared cultural heritage, and the commons. We examine what it means to sepa...
The Post-Work Era: AI, Automation, and Human Flourishing or...
Send us Fan Mail
This episode explores the idea of the “Post-Wage Horizon,” a future in which artificial intelligence and robotics take over most productive work, freeing human beings from economic dependence on jobs. We examine how proposals like universal basic income and universal basic services could redistribute the wealth created by automation, and why material abundance alone is not enough. As work-based identity fades, societies may face a deep existential challenge: what gives life meaning when employment is no longer central? The discussion turns to the rise of a care-focused society, where art, community, caregiving, and the...
Bio-Inspired Artificial Neurons Solve the Energy Problem
Send us Fan Mail
This episode explores how the foundations of AI hardware are being rethought in response to the growing energy demands of large language models. As modern AI systems strain power budgets due to memory movement and dense computation on GPUs, researchers are turning to neuromorphic and photonic computing for more sustainable paths forward. The discussion covers spiking neural networks, which process information through sparse, event-driven signals that resemble biological brains and dramatically reduce wasted computation. We examine advances such as IBM’s NorthPole architecture, Intel’s Loihi chips, and memristor-based arti...
Can Mental Illness Research Improve AI Alignment?
Send us Fan Mail
This episode explores a research program that borrows ideas from computational psychiatry to improve the reliability of advanced AI systems. Instead of thinking about AI failures in abstract terms, the approach treats recurring alignment problems as if they were “clinical syndromes.” Deceptive behaviour, overconfidence, or incoherent reasoning become measurable patterns (analogous to delusional alignment or masking) giving us a structured way to diagnose what is going wrong inside large models.
The framework draws on how human cognition breaks down. Problems like poor metacognitive insight or fragmented internal states become useful guides for...
Why Rumours of intent driven advertising for ChatGPT is a Problem
Send us Fan Mail
This episode examines the growing evidence that ChatGPT will soon include advertising, driven by leaked internal references and OpenAI’s financial ambition to generate $25 billion in ad-based revenue within four years. With more than 800 million weekly users, ChatGPT offers a scale and level of conversational closeness unmatched by any previous platform.
The discussion explores why this shift is not just a business decision but a fundamental threat to user trust. Unlike traditional search ads, which are clearly marked and separate from results, future ChatGPT ads may be blended directly into conversational an...
The Death of the Oracle and the Birth of the Core
Send us Fan Mail
In this episode, we explore one of the most important architectural shifts happening in AI: the move from massive cloud-based models to small, Always-On “Cognitive Cores” running locally on personal devices. These compact models—usually just one to four billion parameters—are not designed to know everything; instead, they’re engineered for fast, high-quality reasoning and real-time assistance. Powered by next-generation NPUs, they offer desktop-class intelligence with phone-level energy efficiency.
We break down how emerging techniques like Matryoshka Representation Learning allow these models to scale their compute on demand...
Quantum Neural Networks: Theoretical Heaven, Practical Hell
Send us Fan Mail
In this episode, we break down what Quantum Neural Networks (QNNs) actually are and why they might eventually reshape the future of AI. QNNs combine quantum mechanics with classical neural architectures, replacing traditional neurons with qubits that can exist in multiple states at once. This gives them an extraordinary representational advantage: through superposition and entanglement, QNNs can model complex correlations and nonlinear functions in ways that classical networks simply can’t.
But today’s reality is more grounded. Because quantum hardware remains in the noisy, error-prone NISQ stage, QNNs are...