Barrett Lyon on Rebuilding Internet Privacy and Infrastructure

The DoxxNet founder discusses mesh networking, tracking, infrastructure operations and his path from DDoS defense to internet art.

BLBarrett LyonInvité
Chaîne
a16z
Publié
1 oct. 2026
Vidéo
45:06
Lecture
5 min
Original
anglais
Ouvrir sur YouTube
En faisant défiler, la vidéo se réduit en bas à droite et continue. Les sous-titres suivent votre langue de lecture.
Langue de lectureDisponible en 2 langues
Cet article n’est pas encore disponible en français ; il s’affiche dans l’original en anglais.
RÉSUMÉ

En 30 secondes

Barrett Lyon talks with Joel de la Garza about DoxxNet, concerns over centralized services and tracking, and lessons from building network infrastructure. Product and security claims are attributed to Lyon; discussion of data-center water use is placed in context.

  1. 1A mesh-network proposalLyon describes encrypted peer-to-peer services built on DoxxNet’s network.
  2. 2Trust and infrastructureHe argues users should examine who owns and operates underlying infrastructure.
  3. 3AI operationsThe founder says the company uses AI to administer a small team’s distributed infrastructure.
  4. 4Water varies by designCooling systems differ; some facilities consume water through evaporation.
  5. 5The Opte ProjectA network-mapping experiment became museum-displayed art.

À retenir

  1. Lyon wants to build encrypted peer-to-peer services on a configurable mesh.
  2. A privacy service requires scrutiny of its operators and infrastructure.
  3. The company says it runs its own infrastructure and local AI operations.
  4. Data-center water use varies with cooling design.
  5. Lyon’s earlier work spans DDoS defense and network visualization.
TEXTE INTÉGRAL

Lire le texte remanié, chapitre par chapitre

Transcription seule
01

An emergency in the air

Joel de la Garza introduces internet infrastructure entrepreneur Barrett Lyon. Before discussing Lyon’s new network, they begin with an aviation emergency that changed how he thought about risk.

Joel de la Garza introduces internet infrastructure entrepreneur Barrett Lyon. Before discussing Lyon’s new network, they begin with an aviation emergency that changed how he thought about risk.

Lyon had started flying between the Bay Area and the Tahoe region to avoid long commutes. After a meeting, his aircraft lost most engine power near Oakland; he declared an emergency and looked for a safe place to land.

He considered the aircraft’s parachute but decided he could reach an airport, where he landed. Only after a firefighter approached did the danger fully sink in. Lyon later sold the plane, feeling the risk was too high.

Lyon recalls several unusual failures occurring together: a broken intake component and problems with both electronic ignition and the backup magneto. It took time to understand the cascading failure.

02

Building DoxxNet

The name DoxxNet goes back to Lyon’s high-school venture selling shell accounts. He retained the domain and later reused it for a project motivated by privacy concerns, censorship and a desire to experiment with network protocols.

The name DoxxNet goes back to Lyon’s high-school venture selling shell accounts. He retained the domain and later reused it for a project motivated by privacy concerns, censorship and a desire to experiment with network protocols.

Lyon argues that much recent internet development has happened above older protocols. He wants to explore mesh and peer-to-peer networking, including approaches to the obstacles created by carrier-grade network address translation.

The speakers discuss content restrictions and age verification. Lyon argues that IP addresses are a poor proxy for personal identity and says network-level restrictions can have unintended effects when geographic classifications are wrong.

03

VPNs, privacy and trust

Lyon describes a conventional VPN as an encrypted connection to an entry and exit point, and questions whether users understand the provider and underlying hosting. His proposed alternative is a broader network with configurable addressing and policy.

Lyon describes a conventional VPN as an encrypted connection to an entry and exit point, and questions whether users understand the provider and underlying hosting. His proposed alternative is a broader network with configurable addressing and policy.

He says DoxxNet provides a mesh with its own domains, address leasing and policy controls, with calling, video, messaging and file transfer built on top. These are the founder’s descriptions of the product, rather than independently verified security guarantees.

According to Lyon, an invitation exchanges credentials so trusted users can connect directly through the mesh without sharing phone numbers or email addresses. He says the design supports large encrypted transfers between peers.

Lyon contrasts this approach with messaging systems that use intermediary servers. Server routing and end-to-end encryption are separate design questions, however: the presence of a server alone does not establish that message content is readable by the provider.

De la Garza recalls the FBI’s ANOM operation as a warning that users must consider who operates a supposedly secure service. Justice Department records describe a covert platform distributed through criminal networks and monitored by law enforcement. [Source: U.S. Department of Justice](https://www.justice.gov/usao-sdca/pr/fbi-s-encrypted-phone-platform-infiltrated-hundreds-criminal-syndicates-result-massive).

Lyon says DoxxNet has separate US and Swiss operations and that he wants its ownership and decisions to be visible. He urges people evaluating a VPN to ask not only about its stated logging policy but also who owns the servers and carries the traffic.

04

Infrastructure and AI operations

He describes building and controlling physical infrastructure as a way to set policy, while acknowledging that carriers and interconnections form another layer. His company says its routing and AI components run on its own equipment.

He describes building and controlling physical infrastructure as a way to set policy, while acknowledging that carriers and interconnections form another layer. His company says its routing and AI components run on its own equipment.

Lyon says the company has roughly 26 sites and about a dozen people involved, with AI handling network and system administration. He argues that a small number of people with access reduces one category of insider risk; these operational and security claims are his account.

05

Data centers and tracking

The conversation turns to data centers. Lyon disputes some online depictions of noise and water use, and both speakers speculate that commercial interests may be shaping opposition; they do not present evidence establishing an organized campaign.

The conversation turns to data centers. Lyon disputes some online depictions of noise and water use, and both speakers speculate that commercial interests may be shaping opposition; they do not present evidence establishing an organized campaign.

Water use depends on the facility and cooling design. The U.S. Department of Energy notes that cooling towers can consume water through evaporation, so a closed internal loop does not by itself establish that a whole facility uses negligible water. [Source: DOE](https://www.energy.gov/cmei/femp/cooling-water-efficiency-opportunities-federal-data-centers).

Lyon’s privacy concern is the aggregation of many small signals across apps and websites. He says combined tracking can reveal routines and life events, and argues that people should have more control over it.

The host adds that increasingly capable AI could analyze collected data at scale. Both speakers worry about the implications for profiling and targeting, while presenting this as a risk rather than evidence about a particular campaign.

06

A career in network defense

Lyon traces his security career to attacks on his high-school servers. As distributed denial-of-service attacks grew, he helped businesses defend themselves and founded Prolexic around shared, large-scale protection infrastructure.

Lyon traces his security career to attacks on his high-school servers. As distributed denial-of-service attacks grew, he helped businesses defend themselves and founded Prolexic around shared, large-scale protection infrastructure.

He later worked on Defense.Net and BitGravity. In his account, video delivery highlighted the cost of transferring large files and the need for infrastructure designed around media access rather than generic downloads.

For DoxxNet, Lyon wants to apply that experience to individuals, not only enterprises. He also recognizes that an internal attack on the network would create its own operational challenge and says privacy does not remove legal obligations.

07

Mapping the internet and art

The conversation shifts to the Opte Project, Lyon’s visualization of internet networks. What began as a bet that he could map the internet grew into images and animations shown in museums, including MoMA.

The conversation shifts to the Opte Project, Lyon’s visualization of internet networks. What began as a bet that he could map the internet grew into images and animations shown in museums, including MoMA.

Lyon describes the surreal experience of attending a museum opening as an artist and the impostor feelings that followed. The project also lets viewers see changes in the network over time, including the rise and fall of companies.

A book about the early Prolexic years recounts efforts to confront cybercriminals. Lyon remembers the strain of protecting clients during outages and a tense trip to Russia; he describes these as personal recollections.

08

Burning Man and connectivity

Finally they discuss Burning Man, the Mind Warrior art car and Lyon’s camp. He credits his partner Pablo with creating the vehicle and describes how shared enthusiasm drew other people into related projects.

Finally they discuss Burning Man, the Mind Warrior art car and Lyon’s camp. He credits his partner Pablo with creating the vehicle and describes how shared enthusiasm drew other people into related projects.

Connectivity on the playa has tradeoffs, Lyon says: satellite access helps organize a camp and resolve practical problems, even if constant outside entertainment can distract from the event. The conversation ends with his hope of reviving a more open internet experience.

Sur la même chaîne

1:00:17EN
a16z· 28 oct. 2026Podcast · 3 voixMark Cuban on Healthcare Prices, AI and the Case for TransparencyInvité : Mark Cuban · Anime : Julie Yoo, Erik TorenbergThe Cost Plus Drugs co-founder discusses contracts, personalized medicine and entrepreneurship with Julie Yoo and Erik Torenberg.AI in healthcarePricingEntrepreneurship+1
56:00EN
a16z· 27 oct. 2026Podcast · 3 voixJoe Liemandt on Alpha School, AI, and Redesigning the School DayInvité : Joe Liemandt · Anime : Erik Torenberg, Gagan BiyaniAlpha School's principal discusses two-hour academics, motivation, life skills, and how AI could change what students build.EducationAI StrategyEntrepreneurship
1:00:39EN
a16z· 23 oct. 2026Podcast · 3 voixAmazon CEO Andy Jassy on the $220 Billion AI BuildoutInvité : Andy Jassy · Anime : Ben Horowitz, Erik TorenbergEven at roughly $220 billion of 2026 CapEx, Jassy says Amazon is still turning away AI demand. He tells Ben Horowitz and Erik Torenberg why AWS could become a trillion-dollar run-rate business, and why inference cost, custom chips, and smaller teams matter more than squeezing old customers.AI StrategyInfrastructure et déploiementSemiconductors+2
1:00:34EN
a16z· 22 oct. 2026Podcast · 4 voixMarc Andreessen, Chris Dixon and Ben Horowitz on AI and the Frontier for FoundersInvité : Chris Dixon, Marc Andreessen, Ben Horowitz · Anime : Erik TorenbergThe a16z partners explain their expanded Frontier mandate and why they think AI is widening the possibilities for entrepreneurs, open models and new media.Venture CapitalAI startups & small teamsAI Strategy+2
55:07EN
a16z· 20 oct. 2026Podcast · 2 voixKaran Singhal on AI, Trust and the Future of Health CareInvité : Karan Singhal · Anime : Jay RughaniOpenAI’s health AI research lead discusses medical evaluations, patient support and the gap between information and care.AI in healthcareAI SafetyAI research
1:01:59EN
a16z· 19 oct. 2026Podcast · 2 voixJean-Denis Greze on Leaving Tax Software to Build an AI AssistantInvité : Jean-Denis Greze · Anime : Alex RampellThe Town founder explains why a growing tax business was not the opportunity he wanted, and how he thinks AI agents could change everyday work.AI startups & small teamsPersonal agentsEntrepreneurship+1