Barrett Lyon Deploys Owned-Metal Mesh to Bypass Cloud Intermediaries

Joel de la Garza, Barrett Lyon

Encrypted communications remain vulnerable to automated AI surveillance unless networks eliminate intermediate application servers by owning physical infrastructure and routing traffic directly peer-to-peer.

AI-assisted summary of Rebuilding the Internet for Privacy | Barrett Lyon on DoxxNet — original source. Check the original for full context.

Source published: 2026-10-01

Key ideas

• Barrett Lyon created DoxxNet to rebuild consumer networking from the physical hardware up, deploying 26 redundantly routed global sites in roughly six months with a 12-person team and zero sysadmins. (21:28)
• Unlike centralized messaging apps like Signal and WhatsApp that route and store traffic through intermediary servers, DoxxNet uses an API-driven parallel mesh where user devices communicate directly peer-to-peer with three to four layers of post-quantum encryption. (14:40)
• Lyon argues that traditional commercial VPNs merely lease third-party bare metal and act as centralized proxies, warning that foreign infrastructure providers and organized crime often compromise data privacy behind 'no-logs' claims. (20:00)
• The operational stack relies on proprietary, self-hosted local AI agents for monitoring and automation, sharing zero telemetry, prompts, or inference metadata with mega-cap commercial AI providers. (21:28)

Detailed notes

## Background Published on October 1, 2026, as an audio podcast hosted by Joel de la Garza and featuring guest Barrett Lyon, this episode examines what it will take to rebuild internet architecture around privacy, direct communication, and reduced reliance on centralized intermediaries. The discussion focuses on Lyon's project DoxxNet, addressing why traditional VPNs fall short, how tracking data is increasingly exploited by scalable AI, and how DoxxNet utilizes an owned bare-metal mesh network with localized AI operations.

## TL;DR Barrett Lyon argues that modern digital communications have stagnated under centralized intermediaries, carrier-grade NAT, and pervasive ad tracking that is now easily weaponized by scalable artificial intelligence. To solve this, Lyon built DoxxNet as an independent, programmable mesh network that enables true peer-to-peer messaging, file sharing, and calling on privately owned bare-metal infrastructure without third-party server exposure. Lyon also reflects on his engineering past, recounting a harrowing in-flight mechanical failure, his pioneering work founding Prolexic and fighting cyber extortionists, and creating internet topology maps now housed in the Museum of Modern Art.

## Atmosphere & style The conversation carries the casual, evocative intimacy of two seasoned cybersecurity veterans sharing war stories and philosophical reflections over a lifetime of infrastructure battles. The mood oscillates between lighthearted geek humor—such as laser tripwire experiments and Burning Man art cars—and gripping, tense narratives about mid-air engine failures and narrowly escaping Russian organized crime rings. Listening feels like sitting in on an insider debrief that balances sober critiques of corporate surveillance with infectious optimism about rebuilding the open, exploratory spirit of the early web.

## Bulletin - `21:28` Barrett Lyon created DoxxNet to rebuild consumer networking from the physical hardware up, deploying 26 redundantly routed global sites in roughly six months with a 12-person team and zero sysadmins. - `14:40` Unlike centralized messaging apps like Signal and WhatsApp that route and store traffic through intermediary servers, DoxxNet uses an API-driven parallel mesh where user devices communicate directly peer-to-peer with three to four layers of post-quantum encryption. - `20:00` Lyon argues that traditional commercial VPNs merely lease third-party bare metal and act as centralized proxies, warning that foreign infrastructure providers and organized crime often compromise data privacy behind 'no-logs' claims. - `21:28` The operational stack relies on proprietary, self-hosted local AI agents for monitoring and automation, sharing zero telemetry, prompts, or inference metadata with mega-cap commercial AI providers.

## Key takeaways - `00:00` Centralized consumer messaging apps like WhatsApp, Signal, and Telegram create the illusion of direct communication while routing all traffic and metadata through central intermediary servers. - `00:00` Pervasive ad-tech tracking combined with scalable, 'coin-operated intelligence' allows cheap LLMs to aggregate disparate telemetry data and construct detailed behavioral profiles of users. - `07:28` Carrier-Grade NAT (CGNAT) and middlebox routing hacks have largely crippled the internet's original peer-to-peer potential, forcing devices to depend on intermediary servers to negotiate connections. - `10:00` Policymakers and automated filtering services frequently treat IP addresses as digital identity, leading to blunt over-enforcement and accidental regional blackouts, such as Cisco Umbrella severing a misclassified carrier. - `12:35` Traditional consumer VPNs provide only rudimentary IP obfuscation over rented servers, whereas DoxxNet operates an independent, programmable parallel mesh network with dynamic routing and policy management. - `14:40` DoxxNet implements transparent public key infrastructure underneath standard user interactions, converting user devices into direct servers to enable multi-gigabyte transfers at line speed without third-party storage. - `17:28` Free VPNs and obscure shell-company privacy providers frequently act as honeypots or covert data-harvesting pipelines, underscoring the rule that users become the product when a service is free. - `20:00` True privacy policy enforcement requires ownership down to the physical bare metal and carrier interconnects, as leasing servers cedes operational control to third parties and host jurisdictions. - `21:28` Operating with local, self-hosted AI rather than a large team of human administrators drastically reduces insider-threat risk by eliminating broad administrative credential access. - `24:15` Public backlash against data center power and water consumption is largely fueled by manufactured online panic; modern facilities are closed-loop, quiet concrete structures that bring economic value to host communities. - `28:07` Blocking non-essential telemetry endpoints like dit.whatsapp.net does not break application functionality, exposing how tech platforms gather behavioral diagnostics purely for persistent tracking. - `31:37` Early DDoS extortion schemes against e-commerce and gambling sites led Lyon to invent cloud scrubbing architectures through Prolexic and Defense.Net, shifting the cost curve against attackers. - `35:28` Automating active reconnaissance techniques during penetration tests led Lyon to create The Opte Project, producing visualizations of global internet topology that entered MoMA's permanent collection. - `40:00` Early cyber defense investigations against Russian extortion rings revealed that online denial-of-service attacks were often run as secondary operations by organized crime syndicates involved in narcotics and human trafficking. - `42:55` While Starlink satellite connectivity streamlines critical logistics and supply operations at Burning Man, ubiquitous suburban-style internet risks eroding the presence and immersion of communal experiences.

## Detailed walkthrough ### `00:00` Centralized Intermediaries and the Stagnation of the Internet

Modern digital communication operates under the illusion of direct connection, but virtually all consumer traffic is routed through centralized third-party intermediaries. Applications such as WhatsApp, Signal, iMessage, and Telegram present an interface that looks like direct person-to-person messaging, yet every packet passes through central application servers whose ownership and surveillance vulnerability raise critical privacy concerns. At the same time, the broader internet is confronting an inflection point driven by what Barrett Lyon terms the arrival of scalable, “coin-operated intelligence,” where decades of accumulated tracking data can now be ingested and weaponized cheaply by advanced artificial intelligence models. Rather than continuing to innovate at the fundamental network layer, internet protocol design has largely ground to a halt, with modern platforms choosing to layer new applications over stale, centralized architectures.

```mermaid flowchart TD accTitle: Centralized Messaging vs Direct Mesh accDescr: Diagram contrasting how centralized messaging apps route traffic through servers versus direct peer-to-peer mesh networking.

subgraph Centralized["Traditional App Architecture"] UserA["Sender Device"] -->|"Encrypted Payload"| CentralServer["Central Server / Intermediary\n(Collects Metadata & Logs)"] CentralServer -->|"Relayed Payload"| UserB["Receiver Device"] end

subgraph DirectMesh["DoxxNet P2P Approach"] PeerA["Peer Device A"] <-->|"Transient Encrypted P2P Connection\n(No Server In Middle)"| PeerB["Peer Device B"] end ```

To break this reliance on intermediaries, Lyon created DoxxNet—a project designed to reconstruct internet infrastructure from the physical hardware up. Instead of depending on conventional virtual private networks (VPNs) or public cloud inference APIs that merely shift data exposure elsewhere, DoxxNet focuses on direct, transient peer communication without intermediary storage or centralized servers. Barrett Lyon emphasizes the root problem confronting digital communications today: “All of those applications disguise themselves as you sending something to somebody else. You're really sending something to a server. The question is like, okay, well, whose servers are those?”

---

`01:14` The Commuter Flight: An In-Flight Emergency over the Bay

Before turning his attention back to networking protocols, Barrett Lyon treated aviation as an everyday engineering solution to Northern California traffic. Long before remote work became normalized during the COVID-19 pandemic, Lyon’s daily commute from the Tahoe region down to the San Francisco Bay Area had deteriorated from a 15-minute drive into a grueling 45-minute slog. To eliminate the friction of ground transit, he earned his pilot’s license and turned a Cirrus single-engine aircraft into his primary commuter vehicle, accumulating over 1,300 flight hours.

Following a partner meeting at Andreessen Horowitz (a16z), Lyon departed San Carlos Airport to head back toward Tahoe. While climbing out over Oakland International Airport, his engine suddenly began sputtering. Initially suspecting water contamination in the fuel supply, he maintained his climb and successfully cleared the Class Bravo airspace boundaries. The moment he cleared the airspace, the aircraft suffered a complete loss of engine power.

Faced with a dead stick, Lyon immediately began scanning the terrain below for an survivable landing site, declared an in-flight emergency with air traffic control, and sent a brief note to his friends: “Hey, guys, I have an emergency. Love you.” Gliding without power, he initially aimed for an open field while debating whether to deploy the airframe's built-in Cirrus Airframe Parachute System (CAPS). He was acutely aware of the morbid paradox surrounding the Cirrus: while regarded as one of the safest general aviation aircraft ever engineered, it frequently registers fatal accidents because pilots hesitate to trigger the ballistic rocket parachute, which destroys the airframe upon deployment and impacts the ground hard on energy-absorbing seats. Spotting Buchanan Field Airport across the mountain ridge in Concord, Lyon opted to glide toward the runway rather than sacrifice the plane to an uncontrolled field touchdown.

---

`04:55` A Cascading Mechanical Failure and Leaving Aviation

Barrett Lyon touched down smoothly dead-stick at Buchanan Field, rolling off the runway directly into transient parking without injury or visible exterior damage. The psychological weight of the near-disaster only hit him when emergency services arrived on the scene. As an oversized airfield crash truck pulled up alongside the parked plane and an emergency responder stepped down to shake his hand, Lyon realized how close he had come to an unsurvivable accident:

> “The guy gets out of that, shakes my hand, and then that's when the fear hit me because I'm like, oh, no. It's the guy scraping my brains off. This is the guy that puts the fire out. That was close.”

```mermaid flowchart LR accTitle: Cascading Cirrus Engine Failure accDescr: Flowchart showing the three-stage cascading mechanical and electrical breakdown that caused the engine to fail.

subgraph Mechanical["Mechanical Breakdown"] V["Valve / Lever Snaps"] -->|"Sucked into Manifold"| I["Air Intake"] I -->|"Drops to Oil Pan"| C["Cylinders 3 & 6 Crushed"] end

subgraph Electrical["Electrical Breakdown"] C -->|"Induces Short Circuit"| E["Electronic Ignition Fails\n(Lost Primary Spark)"] E -->|"Simultaneous Backup Failure"| M["Magneto Fails at High RPM"] end

Mechanical --> Electrical ```

A subsequent mechanical inspection revealed an improbable triple-failure that had cascaded through the powertrain: - A mechanical valve lever broke loose and was sucked straight into the air intake manifold. - The debris dropped down to the oil pan and was crushed directly into cylinders three and six. - The resulting mechanical damage triggered an electrical short inside a newly installed electronic ignition system, completely killing primary spark to half the engine. - The independent mechanical backup magneto, which still functioned well enough at low idle to start the plane on the ground, failed completely under high flight RPMs.

Unwilling to tolerate that compounding level of mechanical risk on regular commutes, Lyon rented a car to complete his trip home that night. He later ferried the repaired Cirrus to Las Vegas, sold it, and resolved that any future aircraft he boarded would require fewer moving parts and professional pilot redundancy.

---

`06:05` The Origins of DoxxNet: From High School Shell Accounts to Protocol Stagnation

The motivation behind Lyon's latest company, DoxxNet, ties directly back to his earliest days in computing. In high school, using his teenage hacker handle “Doxx,” Lyon and a classmate founded the original incarnation of DoxxNet in the era before public cloud infrastructure existed. Renting physical server capacity, the pair ran a profitable business selling terminal shell accounts to a community of around 1,000 active users for $5 a month:

> “As a kid, my hacker name was Doxx. And then I started a business in high school with my high school friend called DoxxNet. It was before cloud, and we were renting server products... selling shell accounts for $5 a month, but we had a thousand users.”

Although Lyon shuttered the original operation after graduating from high school to pursue his professional engineering career, he maintained continuous ownership of the domain name.

In recent years, Lyon observed that the internet's protocol layer had frozen in place. Innovation largely abandoned low-level transport and routing protocols, opting instead to stack ever-more centralized web services and ad-supported trackers directly on top of legacy architecture. This protocol stagnation coincided with the aggressive normalization of internet censorship, pervasive corporate behavioral tracking, and the mass consumer adoption of commercial VPNs. Lyon realized that consumer VPNs serve merely as a superficial patch—routing traffic through a different centralized proxy rather than resolving the core architectural vulnerabilities of the network. He decided to revive DoxxNet as an alternative, privacy-preserving networking layer designed from scratch to defend users against malicious actors, foreign intelligence services, ad-tech tracking networks, and intrusive surveillance.

---

`07:28` The Death of Peer-to-Peer and the Curse of Carrier-Grade NAT

The modern internet has strayed radically from the philosophical vision of its early creators. The early network was championed by free-speech maximalists who envisioned a decentralized, resilient web of interconnected nodes capable of routing around censorship and facilitating open public discourse. Over time, that open topography became concentrated into centralized chokepoints. Lyon highlights how state-dominated telecommunications entities—such as state telecom infrastructure in authoritarian regimes like China—exploit centralized topology as an architectural chokehold to govern the flow of information and coerce the decision-making of entire populations.

Technologically, the edge of the network has never been more capable: consumer devices boast fast multi-core processing, gigabit bandwidth, and powerful local Wi-Fi radios that should theoretically enable robust mesh routing and direct peer-to-peer (P2P) transfers. However, that potential has been crippled by middlebox routing hacks, specifically Carrier-Grade NAT (CGNAT), which prevents end-user devices from directly addressing or connecting to one another without an intermediate server negotiating the exchange. Lyon identifies eliminating this artificial friction as the central technical objective of DoxxNet: “And kind of really one of the biggest stains of the entire internet is carrier grade NAT. It's just destroying applications' ability to do really cool stuff. And erasing that is kind of what I'm trying to go for.”

Joel de la Garza notes that the pressure to monitor, filter, and restrict internet communications is no longer exclusive to foreign autocracies like China or Russia; domestic state governments across the United States have steadily begun imposing strict know-your-customer (KYC) identity verification rules and digital access restrictions on internet services.

`10:00` The Flaws of Network-Level Censorship and Identity

Joel de la Garza and Barrett Lyon examine how internet censorship is quietly encroaching into the United States under the guise of content moderation, health restrictions, and age verification. Lyon points out that policymakers repeatedly make the architectural mistake of treating the network layer as an identity verification mechanism. Because the base internet protocols were never designed to identify who sits behind a screen, assigning regulatory enforcement to network operators results in blunt, broken outcomes. As Lyon puts it:

> “An IP address is not your driver's license.”

Lyon highlights how crude network-level enforcement causes massive collateral damage across the internet. He cites a recent incident where Cisco Umbrella broke connectivity for a major telecommunications carrier after misclassifying the carrier as located in France. France had issued a court order mandating the blocking of over one hundred piracy websites; in response, the filtering logic went “scorched earth” and effectively severed the French footprint, inadvertently catching the misclassified carrier in its blast radius. De la Garza concurs, noting that classification engines regularly fail nuance: an AI companionship and "e-spouse" startup his team evaluated was broadly blocked as adult content by half of the industry's classification engines despite hosting no untoward material. Lyon shares that DoxxNet faced a similar hurdle early on because the double-X in its name and similarity to "doxing" caused it to be automatically flagged by safety filters like Claude before the team cleared the issue through formal registration channels.

- **The IP Identity Mismatch:** Network infrastructure treats policy enforcement like a hammer hitting a nail, confusing a routing address with verifiable human identity. - **The Blast Radius of Geo-Blocking:** Oversimplified regional blocking rules applied by services like Cisco Umbrella can accidentally sever connectivity for entire carriers if geolocation metadata is off. - **Overzealous Filtering Heuristics:** Automated classification engines routinely flag legitimate services—from companionship chatbots to privacy platforms named DoxxNet—as illicit or adult material.

---

`12:35` Beyond Traditional VPNs: Building a Programmable Mesh Network

De la Garza frames the rise of privacy products as a direct response to this "creeping information totalitarianism," questioning why standard consumer virtual private networks (VPNs) no longer suffice. Lyon explains that traditional commercial VPNs offer an extremely rudimentary architecture: operators rent generic server slices from third-party data center landlords, install basic tunneling software, and route encrypted traffic through that single node to mask the user’s public IP address. While this provides an elementary obfuscation layer, it is fundamentally brittle and centralized.

In contrast, DoxxNet treats VPN protocols merely as an on-ramp into an entirely independent, API-driven parallel network. Lyon’s team has engineered a dynamic mesh architecture where nodes, routing paths, and policies can be spun up, configured, or dismantled programmatically within milliseconds. Within this mesh, users can register internal domains, lease mesh or public IP addresses, and enforce custom security policies programmatically via automated agents. Lyon explains that building this programmable substrate allows them to deploy peer-to-peer (P2P) communication services directly over the mesh:

> “Inside, we can create either, you can call it a walled garden or a parallel internet or wherever you want to call it, but we built an entire mesh network.”

```mermaid flowchart TD accTitle: Centralized Relay vs DoxxNet P2P Mesh Architecture accDescr: Diagram contrasting traditional relay through a central server with DoxxNet direct peer-to-peer mesh.

subgraph Traditional["Traditional Centralized App Architecture"] UserA["Sender Device"] -->|"Upload payload"| CentralSrv["Central Application Server / Storage"] CentralSrv -->|"Download payload"| UserB["Recipient Device"] end

subgraph DoxxNet["DoxxNet Parallel Mesh Architecture"] MeshA["Sender Device (Acts as Server)"] <-->|"Point-to-Point Mesh Route (Post-Quantum 3-4 Layer Encryption)"| MeshB["Recipient Device"] end ```

---

`14:40` True Peer-to-Peer Communication and Invisible PKI

Addressing De la Garza’s traditional security assumption—that any device containing a battery or electrical current is vulnerable to surveillance—Lyon explains how DoxxNet enables secure, transient communication without maintaining centralized records. The onboarding process verifies human presence through a lightweight puzzle rather than phone numbers or email addresses, issuing the user a persistent yet rotatable token that can be wiped or replaced at will. When two users connect, the application avoids confusing cryptographic jargon: inviting a friend triggers an automated public key infrastructure (PKI) exchange beneath the user interface, establishing mutual cryptographic trust transparently.

Once trust is established, the devices communicate directly across the mesh without an intermediary server. Lyon contrasts this with standard encrypted messaging apps such as Signal, WhatsApp, iMessage, and Telegram—which he labels "creepy." In those centralized ecosystems, users do not send files directly to one another; they upload data to an intermediary cloud server, which subsequently notifies the recipient to pull it down. That centralized choke point restricts transfer speeds, exposes connection metadata, and caps file sizes. DoxxNet removes the intermediary entirely by converting each user's endpoint into a direct server to their peer:

> “It’s really no middle of man, no server in the middle, and your phone becomes a server to your friend.”

- **Invisible Cryptography:** Users exchange cryptographic certificates seamlessly under the guise of an ordinary "invite friend" prompt, bypassing complex PKI management. - **Uncapped Data Transfers:** Because data travels directly peer-to-peer across the private mesh without resting on a storage server, users can transfer massive files—such as 20-gigabyte archives—at line speed. - **Multi-Layered Security:** Connections run across three to four distinct layers of encryption, engineered to be post-quantum resilient.

---

`17:28` Corporate Governance and the Deception of Free VPNs

The discussion turns to the integrity of the broader privacy ecosystem, with De la Garza pointing out that many commercial and free VPN services operate as honeypots or covert surveillance pipelines. He references international law enforcement stings where criminal syndicates abandoned mainstream tools like Signal in favor of supposedly bespoke "unbreakable" encrypted devices and private apps, only to discover the platforms were covertly orchestrated operations run by the FBI and international police. On consumer app stores, dozens of free VPNs present similar risks by monetizing user surveillance. Lyon warns that zero-cost privacy tools are inherently predatory:

> “If you're not paying for it, you are the product.”

Lyon stresses that technical architecture must be backed by transparent corporate governance and clear legal jurisdictions. To insulate users from single-jurisdiction overreach and subpoena abuse, DoxxNet is split across corporate entities: a primary US company paired with a dedicated Swiss entity governing European operations, with similar regional separations planned internationally. Lyon criticizes the pervasive industry trend of shell companies operating VPNs with anonymous owners, concealed business models, and zero accountability. De la Garza adds that the immediate threats everyday users face are rarely sophisticated three-letter intelligence agencies like the NSA or CIA; rather, the real danger comes from lower-level adversaries and shady network operators, such as the rogue VPNs responsible for siphoning crypto credentials during early cryptocurrency market surges.

`20:00` Bare Metal as a Service and the VPN Underbelly

Joel de la Garza and Barrett Lyon discuss the murky reality behind commercial VPN operations abroad, where state intelligence apparatuses and organized crime frequently blend together. While many commercial VPN providers advertise strict “no-logs” privacy policies, users rarely understand who actually owns the underlying hardware. Barrett points out that if a brand leases its servers from foreign third parties, those third parties dictate what happens to the data regardless of the brand’s customer-facing marketing. Lyon realized that the only way to avoid the compromised "bare metal as a service" ecosystem was to construct DoxxNet’s hardware footprint from scratch across the globe.

As Lyon explains, control over privacy begins at the physical layer: “Learning about that has been difficult because I personally was not up to be building infrastructure all over the world as rapidly as I could. But I ended up realizing that that was, you know, if it's not your equipment, you can't set policy. It's just you can't.” Beyond the server cage itself, the next boundary is carrier interconnects, which required immense engineering effort to bypass leased infrastructure risks. Joel highlights that this end-to-end approach sets DoxxNet apart from competitors: they own the entire operational stack from application software down to the physical wire.

```mermaid flowchart TD accTitle: DoxxNet Infrastructure Control Model accDescr: Diagram illustrating DoxxNet's full-stack ownership versus traditional leased VPN architecture.

subgraph Leased["Traditional VPN Model"] A["VPN Brand Marketing<br/>('No Logs' Policy)"] --> B["Leased Bare Metal<br/>(Third-Party Hosts)"] B --> C["Foreign Jurisdictions &<br/>Organized Crime / Intelligence"] end

subgraph Owned["DoxxNet Model"] D["Software Layer<br/>(Peer-to-Peer Routing)"] --> E["Private AI Agents<br/>(Local Automation)"] E --> F["Owned Physical Metal<br/>(26 Global Sites)"] F --> G["Direct Carrier Interconnects"] end ```

`21:28` Eliminating the Insider Threat with AI-Driven Network Operations

Joel recalls that Barrett was deeply skeptical of large third-party AI and inference providers when they first met, prompting Barrett to build a completely proprietary, self-hosted AI stack. Despite the acute global GPU shortage, power constraints, and rising anti-data center sentiment, DoxxNet deployed 26 redundant routing sites globally in roughly six months. Incredibly, the entire company consists of just 12 people and employs neither a traditional network administrator nor a systems administrator; automated AI systems handle the operational load while Barrett oversees them.

This lean operational model dramatically shrinks the corporate blast radius. While insider threats remain among the most devastating vectors for catastrophic cyber intrusions, DoxxNet eliminates that exposure by not having hundreds or thousands of employees with backend credentials. The AI agents run strictly inside a secure internal environment that never exports context, prompts, or telemetry back to mega-cap AI providers.

Physical security is monitored via AI-enabled cameras that trigger alarms on unauthorized motion. Barrett jokes that they even evaluated mounting tripwire lasers in their data center cages for a touch of cinematic flair: “I'm like, why not? It'd be kind of fun. You've got to have the one data center that you take people to to impress them, right? Like the Switch data centers in Las Vegas. They've got the Jeeps with the.50 cal machine guns on them.” They acquired light-actuated switches to trip alarms if the beams were broken, but the plan foundered on an unglamorous technical reality: data center lights stay on continuously, and cleanrooms lack the floating dust or particulates needed to make laser beams visible without blowing forbidden smoke into the facility.

- DoxxNet maintains 26 globally distributed, redundantly routed sites built in approximately six months. - The company operates with a headcount of only 12 people, relying on proprietary local AI rather than dedicated sysadmins or network engineers. - Zero inference or routing metadata is shared with external commercial AI providers.

`24:15` Dispelling Data Center Myths and Manufactured Panic

The conversation turns to the mounting public backlash against data centers, which Barrett and Joel view as fueled by rampant disinformation. Mainstream media reports and viral videos frequently depict data centers as destructive entities that guzzle entire municipal water reserves and emit unbearable shrieks into surrounding communities. Barrett finds the hysteria absurd, noting that data centers are closed-loop systems that operate quietly, with sounds contained entirely inside the facility walls; videos purporting to show loud exterior noise often splice in dubbed indoor recordings.

Joel ranks data centers alongside commercial airliners as the most sophisticated machines humanity has ever constructed, while Barrett observes that from the outside, they are simply boring, quiet concrete buildings surrounded by parking lots. When Joel asks if Barrett has ever encountered protesters during his recent visits to over twenty data centers, Barrett laughs that he has seen nothing except bunny rabbits resting in the open grass fields around the massive data center parks in Northern Virginia.

Far from harming local populations, these hubs have turned host communities into some of the wealthiest counties in the United States, creating lucrative technical jobs and vastly improving local internet latency. Barrett addresses the rumor that data centers "drink water," pointing out that myths often arise from deliberate misinterpretations—such as the Switch facility near Reno, Nevada, which built an artificial lake purely as a reserve basin in the rare event that closed-loop chillers need to be drained and refilled. Lyon warns that public policy is being warped by cynical online narratives: “The future of our tech is needing to build this infrastructure now. And if we're crippling it, then we're making bad decisions. And we're making bad decisions on Instagram posts. Or TikTok.”

`28:07` Commercial Surveillance, Ad Trackers, and Bulk Profiling

Joel draws a parallel between public panic over intelligence agencies like the NSA and the far more pervasive, sophisticated tracking deployed every day by commercial advertisers. Intelligence agencies now routinely bypass legal search constraints by purchasing location and behavioral profiles directly from commercial data brokers. Barrett highlights how pervasive this tracking is inside everyday consumer hardware and applications: WhatsApp relies on endpoints like `dit.whatsapp.net`, and Apple constantly collects diagnostic metrics under the banner of "Apple telemetry." While tech companies frame these signals as harmless diagnostic metadata, they function in practice as persistent trackers. Lyon notes that blocking these telemetry endpoints breaks neither the apps nor the websites, proving they are non-essential for actual service delivery.

Barrett recalls a conversation with a wealthy, controversial figure who dismissed privacy concerns by asking what harm could possibly come from a corporation knowing he bought underwear on Target's website. Barrett argues that focusing on a single, isolated transaction misses the entire point of modern behavioral surveillance:

> “And then I'm like, well, it's not really that one. It's all of them on every different app that you're using or every different website. And it accumulates and then it's being stored by who knows who, where and being sold. And then when you aggregate all those data points together, you could know if you're a coffee drinker, what time you wake up in the morning, what your income is, where you shop, how often you purchase things. If you're preparing for the birth of a child or perhaps for a divorce or some other big life event. So turning that off doesn't bother me at all.”

`30:00` Reclaiming the Old Internet from Coin-Operated Intelligence

Barrett Lyon and Joel de la Garza discuss how consumer privacy has degraded under pervasive commercial surveillance and automated tracking. Lyon explains that stripping away tracking mechanisms is not just an application-level feature for DoxxNet, but a foundational network architecture goal intended to restore open creative connectivity without surveillance by advertisers or governments. De la Garza observes that the world has hit a major inflection point driven by scalable, automated machine learning models analyzing mass data exhaust. He characterizes this modern threat environment as “coin-operated intelligence at scale,” comparing it to infinite monkeys equipped with infinite typewriters capable of synthesizing mass telemetry.

```mermaid flowchart TD accTitle: Transition from Data Extraction to Agentic Peer Network accDescr: Pipeline showing mass telemetry dumped into LLMs contrasted against DoxxNet decentralized architecture.

subgraph Surveillance["Pervasive Tracking Model"] direction TB T["Ad Trackers & Mass Telemetry"] --> LLM["Coin-Operated LLM Analysis"] LLM --> Target["Automated Targeting & Mass Coordination"] end

subgraph DoxxNetModel["DoxxNet Architectural Countermeasure"] direction TB Mesh["Parallel Mesh Network"] --> Local["Deconstructed Tracking / Local AI"] Local --> Free["Unmonitored Peer-to-Peer & Agentic Network"] end ```

Lyon points out that feeding vast swaths of internet telemetry into large language models makes intrusive behavioral profiling trivially simple: “Yeah, you can take all that and just dump it into an LLM and say, tell me about this person.” He emphasizes that dismantling ad tracking would tangibly improve everyday life, noting that DoxxNet seeks to recreate the exploratory vibe of the early web while building out protocols that do not yet exist. When de la Garza remarks that the internet originally succeeded because it was free and thanks Lyon for trying to make it free again, Lyon jokes back, “Trying. But you have to pay me.”

- De la Garza and Lyon note that automated AI inference can orchestrate coordinated movements across platforms and run targeted intelligence against critical infrastructure and data centers. - Lyon argues that consumer web tools must deconstruct tracking entirely, replacing centralized surveillance silos with open environments for creative experimentation.

---

`31:37` From High School Defenses to Prolexic, Defense.Net, and Big Gravity

Reflecting on their shared history in the early 2000s security ecosystem, de la Garza recalls analyzing early distributed denial-of-service (DDoS) botnets like Stacheldraht, Tribe Flood Network (TFN), and TFN2K. Lyon traces his journey in DDoS mitigation straight back to the original DoxxNet servers he ran in high school. Facing constant attacks, he developed survival tricks, documented them in defensive guides, and presented his findings at SANS. Once those materials were indexed by Google, companies began calling him directly for help navigating high-stakes extortion schemes, such as competing watch retailers taking each other offline during Christmas sales or gambling platforms getting blasted dark right before the Super Bowl.

```mermaid flowchart LR accTitle: Barrett Lyon Infrastructure Evolution accDescr: Flowchart tracing Barrett Lyon ventures from high school servers to modern consumer networks.

subgraph EarlySecurity["Early Security Defense"] DoxxEarly["High School DoxxNet"] --> SANS["SANS DDoS Guides"] SANS --> Prolexic["Prolexic (Acquired by Akamai)"] Prolexic --> DefNet["Defense.Net (Acquired by F5 Silverline)"] end

subgraph ScaledMedia["Bandwidth & Scale"] DefNet --> BigGrav["Big Gravity (Video CDN)"] BigGrav --> DoxxModern["DoxxNet (Consumer & Agentic Mesh)"] end ```

Recognizing that individual businesses could not afford the massive bandwidth pipes and hardware required to absorb large-scale attacks, Lyon realized that DDoS protection had to be pooled into a shared, leased architecture. This insight produced Prolexic, one of the earliest cloud-based infrastructure security services, which was later acquired by Akamai and still retains the Akamai Prolexic branding. Lyon followed this with Defense.Net, which was later acquired and turned into F5 Silverline. Driven by an affinity for massive scale and high bandwidth, he also founded Big Gravity, one of the first video-centric content delivery networks (CDNs). Big Gravity built optimization technologies that eliminated progressive-download bandwidth waste when users skipped into the middle of video files. Now, Lyon is channeling those decades of carrier-scale networking into a consumer-facing, agent-ready network.

> “Now it's kind of like twisting it. Instead of doing this for corporations, we're doing it for people.” — Barrett Lyon

- Early commercial DDoS attacks were driven by cutthroat commercial extortion, timing traffic floods to critical retail windows and live sporting events. - Lyon built Prolexic around high-capacity leased scrubbing centers, pioneering enterprise cloud security infrastructure. - Big Gravity solved bandwidth inefficiencies in video delivery by ending full progressive file downloads when viewers scrubbed video playback. - The new DoxxNet network incorporates agentic networking primitives designed to handle both human users and autonomous bots. - Lyon acknowledges that privacy networks still face internal abuse threats: “The truth is, even if you have privacy, there's still laws and you need to. At least follow them as much as you can.”

---

`35:28` Mapping the Entire Internet and the MoMA Permanent Collection

De la Garza shifts to Lyon’s artistic career, noting that Lyon created pieces that hang in the permanent collection of the Museum of Modern Art (MoMA) in New York. Lyon explains that while studying philosophy, aesthetics, and theoretical AI in college, he worked as a penetration tester. To scope target systems, he routinely drew network topology maps. He soon automated the process and, during lunch one day, impulsively made a $50 bet that he could apply the exact same active reconnaissance technique to map the entire internet. Shortly after making the bet, he was struck by a car while riding his bicycle. Unable to walk and suddenly forced into a slow pace of life, he sat down and wrote the mapping code for what became The Opte Project.

```mermaid flowchart TD accTitle: The Opte Project Origins accDescr: Flowchart showing the creation and exhibition journey of Barrett Lyon Internet mapping project.

PenTest["Pen Testing Network Topology"] --> Bet["$50 Lunch Bet to Map Internet"] Bet --> Accident["Bicycle Accident / Convalescence"] Accident --> OpteCode["Opte Project Visualizations & Code"] OpteCode --> Museums["MoMA, Boston Museum of Science, Global Tours"] OpteCode --> VintCerf["Dialogue with Vint Cerf"] ```

The resulting visualizations and animations captured global attention, showing visual histories of internet evolution where viewers could watch dot-com companies flourish, fail, and crash in network routing space. The work was displayed at the Boston Museum of Science, traveled across international exhibits, and earned two spots in MoMA’s permanent collection, while also giving Lyon the opportunity to discuss internet topology with internet pioneer Vint Cerf.

De la Garza notes that Lyon managed two of the rarest achievements: founding successful technology companies and achieving museum recognition as an artist. Lyon admits that being a successful artist felt far harder than starting a tech business. He recalls attending the celebratory opening at MoMA with drinks, a DJ, and exclusive tours of private MoMA properties, only to crash emotionally the following morning:

> “The next day, I kind of had this imposter syndrome where I just felt super depressed. I'm like, wow, am I ever going to experience what an artist feels like ever again? Because this was kind of like accidental, you know?” — Barrett Lyon

- Lyon holds a philosophy degree with an emphasis on aesthetics and early theoretical artificial intelligence. - The Opte Project mapped internet nodes by automating network discovery techniques originally used to case environments during security audits. - The project documents internet routing shifts over time, visually capturing significant moments of corporate infrastructure growth and collapse.

---

`39:15` The Miami Boat Webcam and Uncovering the Past

Bringing the conversation to current creative experiments, de la Garza and Lyon talk about a book written about Lyon’s early cybersecurity battles. Lyon admits that he frequently forgets the book even exists until strangers bring it up. He shares an anecdote about running a 24-hour, AI-powered webcam that monitors boat traffic in Miami. The stream has run continuously for almost a year and developed an active niche chat community.

A community member in the webcam chat recently connected the camera's host to Barrett Lyon's real-world identity. The user announced in the stream chat that they had just finished reading the published book about him, catching Lyon completely off guard. When de la Garza asks what the book covers, Lyon recalls that it chronicles the beginnings of Prolexic, early distributed cyber attacks, and the cyber extortion investigations from that era.

`40:00` Tracking Cyber Criminals in Moscow

Barrett Lyon recounts the high-stakes battle against aggressive Russian cyber extortionists who targeted and damaged his infrastructure clients. The operational toll was severe, with entire holidays derailed just trying to mitigate distributed denial-of-service disruptions and client fallout. As Barrett puts it:

> “I mean, we had Christmases destroyed because we were just dealing with customers and network outages.”

Joel de la Garza emphasizes that these perpetrators were far from mere mischievous hackers—they operated as hardcore organized criminals engaged in serious illicit enterprises like human trafficking and narcotics smuggling, running cyber extortion as an aggressive sideline.

The confrontation culminated in a trip directly to Moscow. While Barrett was on the ground, the situation escalated until unmistakable indicators warned him that his safety was compromised, prompting a hasty exit from the country. The saga eventually resulted in the perpetrators being sentenced to roughly eight years of hard labor in prison, a wild chapter chronicled in detail in a book. Looking back, Barrett reflects on how surreal the danger felt only in hindsight: living through it seemed almost mundane at the time, but looking back reveals just how extreme it really was.

- **Criminal activities:** Cyber extortion, DDoS assaults, narcotics trafficking, and human trafficking. - **Consequences:** The perpetrators received approximately eight-year hard-labor sentences. - **Investigation aftermath:** Barrett and his team had to abruptly flee Moscow after receiving urgent signals that their lives were at risk.

---

`41:15` Building the Mayan Warrior at Burning Man

Shifting from cybersecurity war stories, Joel brings up Burning Man, noting that Barrett is heading back for roughly his fourteenth burn after missing a couple during the pandemic. The conversation turns to Barrett's involvement with the globally recognized art car, the *Mayan Warrior*. While his creative partner Pablo designed the physical car, Barrett founded and organized the accompanying Mayan Warrior Camp, joining forces to pilot what started as an experimental passion project.

```mermaid flowchart TD accTitle: Mayan Warrior Evolution accDescr: Progression of the Mayan Warrior from a desert passion project to a global touring phenomenon and its eventual retirement. subgraph Origins ["Playa Origins"] A["Pablo creates Mayan Warrior art car"] B["Barrett Lyon founds Mayan Warrior Camp"] A --- B end

subgraph Expansion ["Global Phenomenon"] C["Traveling superpower soundstage & laser production"] D["Worldwide ticketed events & creative influence"] Origins --> C --> D end

subgraph Evolution ["Current Chapter"] E["Soundstage retired from the playa"] F["Pablo introduces two standalone art installations"] D --> E D --> F end ```

The installation grew far beyond the Black Rock Desert, morphing into an international sensation that drew massive crowds and sold out independent venue shows worldwide. Barrett explains how creative energy spreads:

> “What I was thinking about this morning is when you're passionate about something, it kind of creates that aura around you, and then other people pick up on it.”

That project inspired countless other artists and builders across the community. However, Barrett shares that an era has concluded: for the first time, they are not bringing the traveling soundstage car back to the playa, and Barrett will not be presenting his signature laser display. Instead, Pablo is bringing two separate, standalone art pieces to Black Rock City.

---

`42:55` Starlink, Radios, and Connectivity on the Playa

Joel and Barrett discuss the changing culture of Black Rock City following the widespread availability of Starlink satellite internet on the playa. Barrett acknowledges the controversy, noting that there are reasonable arguments on both sides. From the perspective of camp operations, real-time data connectivity solves critical logistical nightmares, such as instantly coordinating water refills for art cars, tracking supplies, or preventing newcomers from getting stranded in long lines at the gate.

At the same time, Barrett highlights the classic alternative: his crew still relies on high-powered Motorola radios and shortwave setups, which carry a playful, nostalgic energy while keeping essential camp communications alive. While he embraces connectivity for pragmatic operations, he draws the line at treating the desert like a suburban living room:

> “If you're using it for anything other than like local communications then maybe it's like the wrong wrong use, but it has changed a bit.”

Joel agrees that attendees do not travel out to Burning Man to "Netflix and chill," cautioning against letting ubiquitous broadband erode the immediacy and immersion of the event.

---

`44:03` Wrapping Up DoxxNet and Old Internet Vibes

Closing the episode, Joel ties Barrett’s history of following deep technical rabbit holes directly to his work building DoxxNet. Praising Barrett's lifelong habit of pursuing genuine obsessions regardless of where they lead, Joel expresses his gratitude as an active, satisfied DoxxNet user:

> “It's good to get that old internet vibe back.”

Barrett responds with humility and excitement, replying, “Yeah, well, it's just getting started. And thanks for trusting me to help build it.”

Joel finishes by urging listeners to share DoxxNet with friends and family, subscribe across podcast platforms, and follow a16z’s Substack and social accounts, followed by standard legal and investment disclosures concerning Andreessen Horowitz and its affiliated portfolio investments.

## Notable quotes > `10:00` An IP address is not your driver's license. — Barrett Lyon

> `20:00` if it's not your equipment, you can't set policy. — Barrett Lyon

## Mentioned Barrett Lyon, Joel de la Garza, DoxxNet, Andreessen Horowitz (a16z), Signal, WhatsApp, iMessage, Telegram, Cirrus, Cirrus Airframe Parachute System (CAPS), Buchanan Field Airport, Oakland International Airport, San Carlos Airport, Cisco Umbrella, Prolexic, Defense.Net, Big Gravity, Akamai, F5 Silverline, The Opte Project, Museum of Modern Art (MoMA), Boston Museum of Science, Vint Cerf, Mayan Warrior, Burning Man, Starlink, Switch, Apple

## Video description a16z’s Joel de la Garza sits down with DoxxNet founder Barrett Lyon to discuss what it would take to rebuild the internet around privacy, direct communication, and less dependence on centralized intermediaries. Barrett explains why he thinks traditional VPNs only solve part of the problem and how DoxxNet is building a parallel mesh network where users can communicate peer-to-peer, transfer large files, make calls, and message without routing those interactions through a central application server. He also explains why the company owns its infrastructure and runs its AI systems locally rather than relying on third-party inference providers. doxxnet_assembly They also get into the growing amount of tracking embedded across the internet, what happens when that data can be analyzed by increasingly capable AI, and why Barrett believes the underlying protocols of the internet are due for another wave of experimentation. Along the way, he shares lessons from decades of building internet infrastructure, from fighting early DDoS attacks to mapping the internet and building global networks.

[Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/f388d562-3c7e-4342-ba98-05729b8c29e8/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=f388d562-3c7e-4342-ba98-05729b8c29e8&feed=JGE3yC0V)