About
I work as a generative action researcher and applied technology ethicist, designing and…
Articles by Mathew
Activity
3K followers
-
Mathew Mytka reposted thisMathew Mytka reposted thisLast week I asked if anyone was still interested in decentralized AI. Turns out... yeah! So here's lecture 1: Decentralized AI From Scratch (Part 1: My First P2P AI) We build a peer-to-peer AI from scratch in about 50 lines of Python. It runs on your laptop, answers your friends' WhatsApp messages using your local data, and begins to address the privacy / prompt-injection problem through user-specific context management. This is the first lecture in a new series called "Decentralized AI from Scratch," building on code/concepts from Karpathy's zero-to-hero but adding cryptography and distributed systems concepts which enable decentralized governance. Future lectures will upgrade this 50-line prototype piece by piece, adding things like homomorphic encryption, federated learning, secure enclaves, ZK proofs, differential privacy, etc. Starts simple, stays simple as long as it can, and will explain things using simple terms as we go along until we have a full-stack decentralized AI system. Blog post: https://lnkd.in/er_u8XdE Video: https://lnkd.in/eQ582BjG Code: https://lnkd.in/eRa_hWdX
-
Mathew Mytka reposted thisMultidisciplinary Association for Psychedelic Studies (MAPS)
Multidisciplinary Association for Psychedelic Studies (MAPS)
Mathew Mytka reposted this83 years ago today, a Swiss chemist named Albert Hofmann stepped out of his laboratory in Basel, Switzerland, climbed onto his bicycle, and rode home accompanied by his lab partner. He’d just ingested 250 micrograms of a compound he’d first synthesized five years earlier — LSD-25. That ride became the world’s first LSD experience✨🚲 -
Mathew Mytka reposted thisYou can never read too much fiction.✅ But the purest of pleasures lies in writing our story of tomorrow.💯 I invite all #InnoMinds to enter Metagov’s Protopian Prize Fiction Contest.🙏 ▶️ http://protopianprize.com Dream big & let’s #FreeTheFuture — together!🫂 #LLAP🖖Mathew Mytka reposted thisWrite the story of humanity’s future. To create a better world, we need to first imagine it – and the work required to get there. The Protopian Prize is a fiction contest inviting artists, storytellers, and dreamers to share their vision of people working toward liberatory futures, meeting obstacles, and making real change. “Protopian”—a word coined by Kevin Kelly, one of our contest's judges—means an achievable, optimistic future characterized by continuous, incremental progress rather than revolutionary leaps or a static, perfect state. Protopian stories imagine a future that is neither flawless nor catastrophic, but instead workably better than today. It’s about plausible progress rather than perfection or collapse. Contest submissions are due on July 31, 2026, and we anticipate awarding prizes in September 2026. There are two prize categories selected by judges: The Public AI category invites us to imagine AI development, policy, and tools, built for—and following basic principles of— public benefit. Stories might address ethical training, sustainable maintenance, creator-friendly implementation, collective rather than corporate ownership, and/or new uses that complement rather than replace human strengths. The Democratic Futures category invites us to imagine institutions, tools, and organizational methods that expand and enhance public autonomy and shared control over societal futures. Stories might address inclusive community-building, decision-making processes, better delivery of public goods, and/or new techniques for combating disinformation, surveillance, and authoritarian constraints. Imagining clear, concrete goals is a common barrier to movements for protecting democracy and advancing civic good. Researchers at Democracy Funders Network and Democracy 2076 have found that even those deeply involved in such efforts understand far more about the negatives that they’re fighting than the positives that they fight for. Stories about what it would look like to work for, and achieve, those positives make it easier to do so. Stories can improve communications, making it easier for organizations to describe what they’re for as well as against. They can support effective planning, making it easier to lay out specific goals and actions to reach them. They also contribute to engagement and participation, by making positive futures feel more plausible and providing an alternative to doomerist learned hopelessness. Find out more at http://protopianprize.com/
-
Mathew Mytka reposted thisMathew Mytka reposted thisOkay, this is pretty amazing. Did you know that olive trees only live for about a hundred years on their own? And they grow into bushy shrubs if left uncared for? Here is an olive tree that is MORE THAN 2000 YEARS OLD because it has been continuously pruned and cared for by more than 20 generations of human beings. They extended its life and created something truly incredible to behold. We visited this tree in the north of Tunisia today. It was so cool!
-
Mathew Mytka reposted thisI'm so. tired. of hearing that we "don't have information" about data center emissions. The numbers definitely exist, but the Big Tech companies building and operating data centers don't want them to see the light of day (and potentially be scrutinized by policymakers and users). What we need to change this is more transparency and more oversight. But honestly -- I think we're seeing the tide turn when it comes to pushback against data center development and the opacity of the whole AI supply chain.Mathew Mytka reposted thisThe growing AI backlash is largely a reaction to the lack of transparency from the companies developing and deploying this technology. An article released today by The Guardian highlights a moment that's emblematic of this opacity: in 2024, US tech companies (usually through trade groups) successfully pressured the European Commission to move away from publishing environmental metrics on individual data centers. The companies cited confidentiality to prevent this transparency, a justification we hear repeatedly at different levels of the AI value-chain (including models, hardware, infrastructure). The net effect is an information vacuum where estimates and misinformation creeps in - undermining the very growth they're trying to promote. If AI is going to scale to the levels that these companies envision, transparency is not optional.
-
Mathew Mytka reposted thisMathew Mytka reposted thisIf an AI agent can call your APIs, run your code, and spend your money, the language it accepts becomes a security boundary. Today these messages are in natural language or unrestricted JSON schemas. No filter or guardrail can catch every deviant message that may cause unexpected computation. Not because we haven't built a good enough filter. Rice's theorem says that every non-trivial semantic property of an arbitrary program is undecidable. So we stopped building better filters and built a different kind of language. CBCL is a formally verified control plane for agent communication. Every message, including messages that extend the language itself, parse in bounded time on a deterministic pushdown automaton. "Is this message safe to process?" always has a computable answer. Agents grow their vocabulary on the fly. They define new message types, teach them to peers, and adopt them without a central registry. Three invariants enforce this, machine-checked in Lean 4: no recursion, static resource bounds, immutable core. The parser binary is extracted directly from the proofs, correct by construction, not by testing. The executable *is* the proof. Paper accepted at LangSec Workshop (IEEE S&P 2026). Implementation, proofs, and draft IETF Internet-Draft open source under Apache 2.0. Preprint in the comments. Thanks to Ric Richardson and the Office for Innovation team for championing this effort. I'll be in San Francisco around mid-May. If any of this resonates, reach out! #LangSec #MultiAgentSystems #FormalVerification #AIAgents #CyberSecurity
-
Mathew Mytka shared thisCurrent AI safety instruments the model. It does not instrument the coupling. Nearly a billion nervous systems now interact with frontier LLMs every week. Their physiological states during those interactions are invisible to alignment work, interpretability work, and evals. That gap has a name now. Preprint on cross-substrate coupling dynamics: https://lnkd.in/g7fptUhB If you're working on sycophancy, AI-induced dysregulation, "AI psychosis", or user wellbeing in LLM interaction, I'd like to compare notes. DM open. #somaticAIsafety ##AIethics #entangledcognition
-
Mathew Mytka reposted thisMathew Mytka reposted thisA lot of the biggest unsolved problems in AI are actually cryptography problems. That's what Andrew Trask, co-founder of OpenMined, argues in our latest episode. Listen to hear how OpenMined is building tools to let researchers learn from sensitive data across institutions without ever having to share the underlying data itself, opening the door to collaborations that just aren't possible today. Andrew shares a vision for a future where better data governance could accelerate breakthroughs in everything from healthcare to financial systems. Available below and wherever you get your podcasts! https://lnkd.in/gPjgSJwj
-
Mathew Mytka reposted thisMathew Mytka reposted thisToday we announced Thunderbolt, an AI client built on the same principles that have guided Mozilla for decades: user control, open standards, and individual empowerment online. As AI becomes central to enterprise operations, organizations face a critical choice: depend on proprietary services that create strategic risk, or build on open infrastructure they control. Thunderbolt gives organizations full control over their data, models, and workflows. Self hostable. Model flexible. Open source. For organizations that need a complete sovereign AI stack, Thunderbolt integrates natively with @Haystack by deepset, powering enterprise grade agents, RAG, and workflow orchestration. Learn more: https://lnkd.in/g_-uQDwg
Mathew Mytka replied to a comment
5d
Hugo O'Connor this is real. I have a heap of anecdotal evidence on that one from many peeps and the broader sweep of commentary online. Not just the dopamine loops. But the integration debt from high frequency and high volume of information + managing agents and decision fatigue. I'd be confident in the assertion that there is a correlation between deterioration in sleep quality after sustained use.
Mathew Mytka replied to a comment
5d
Elizar Mytka thanks for the nudge sis. Just went through the white paper, Privacy Policy and EULA. Michelle Williams had actually mentioned Nicole and the work of intruth last year to me :) Given the adjacent research I'm doing, and the philosophical and tech ethics orientation I come from I've def got many questions the white paper does not answer.
That aside major kudos to you Nicole on the launch!! I know there's some hard work and soul searching involved, particularly when navigating the the ethical tensions of the entrepreneurial journey to bring a product to market like this. Happy to share some feedback and perspective to you and the team when the time is right.
Mathew Mytka commented on a post
6d
We need not replace politicians with AI but we sure could do with the next version of the Flux Party and should evolve to do away with politicians. Flux Party was I involuntarily deregistered in Aus just when it was getting some momentum. This paper is probably of relevance when thinking about agentic systems augmenting the complex collaboration and coordination needs of our civilisation https://lpm.media.mit.edu/agentic_draft.pdf
-
Mathew Mytka liked thisI should make a 300+ co-authors list for this book - Tiffany and her family would be in there for sure. So many years sitting round their table in Bre that they changed my accent! Tiff and her Mum got a nose for wrong story.Mathew Mytka liked thisBooks have always meant more to me than just reading. Growing up without much, my mum made sure I still had access to adventure, culture and bigger worlds through books. It was one of the greatest gifts she gave me, and it has stayed with me ever since. This year I set myself a reading challenge and I’ve finished eight books so far. I’ve just started Right Story, Wrong Story by Tyson Yunkaporta. Tyson has also been a mentor/brother in my life and helped shape some of my early thinking, so returning to his work feels personal as well as intellectually rewarding. What I value most in his writing is the invitation to question the stories we inherit, repeat or accept without even realising. Because the stories we leave unexamined often shape us the most. That feels relevant not only in culture, but in leadership, systems and the way we relate to one another. Would love to hear what others are reading this year. #books #reading #learning #TysonYunkaporta #RightStoryWrongStory #leadership Tyson Yunkaporta
-
Mathew Mytka liked thisMathew Mytka liked thisI'm stepping into a new role at Relevance AI—and we're hiring my replacement. Almost a year ago, I joined Relevance and was given an incredible opportunity: build our support function from scratch and step into my first management role. Since then, our team of three has built processes and AI workflows that autonomously resolve 80%+ of tickets, maintained high CSAT while scaling support coverage, hired two exceptional support specialists (Niamh Byrne and Jordan Cunningham), shipped product fixes and features using Claude and Cursor, and created educational content while going live with customers regularly. 🔮 I continue to believe that the support function is evolving, not disappearing in the AI age. Support teams can be educators, ship code to benefit their customers, and help build Community. Peer-to-peer learning, shared use cases, and collective AI fluency matter more than ever. As Relevance scales, we need dedicated leaders for both Support and Community. I've been given the opportunity to choose where I want to take my career next. Support has been my home for over half a decade. It's given me opportunity after opportunity, and I've loved growing in this craft and evolving what it looks like in the AI age. But it's time for me to step away and go all in on something new—leading our Community motion at Relevance. 🌱 There's a lot more to come on this front: my why, the vision, and how we're scaling the Relevance Community. 🪜 But right now, my focus is on finding the right person to take my team forward and scale it globally. We're hiring a Head of Support in Sydney who will lead and scale our support team globally, inherit two of the best direct reports a first time manager could hope for, and take over one of the most AI-forward support operations in the world. I care deeply about this team and want to make sure we find the right fit. If this opportunity excites you, and we've worked together before, let's chat. 💬
-
Mathew Mytka liked thisI've been testing Kimi K2.6 today. It's *very* good. Definitely "good enough", a post-Watershed model. Gave it a hard problem - we both learned that I'd given it an effectively impossible one. Made an excellent attempt. Rob Manson Andreas Antoniades Snow x Hugo O'Connor Viveka Weiley 🦉 Tony Parisi Robert TercekMathew Mytka liked thisKimi (Moonshot AI) is here with new state of the art open model, Kimi K2.6. Kimi K2.6 features state-of-the-art coding and agentic capabilities. It's available on Ollama's cloud! Try it with OpenClaw: ollama launch openclaw --model kimi-k2.6:cloud Try it with Hermes Agent: ollama launch hermes --model kimi-k2.6:cloud Try it with Claude Code: ollama launch claude --model kimi-k2.6:cloud Model page https://lnkd.in/gPwhuT8a More integrations https://lnkd.in/gzAYAuXw
-
Mathew Mytka liked thisMathew Mytka liked thisJ.P. Morgan just did something scientists couldn't. They made climate tipping points expensive enough to care about 💰. Not in a "this is bad for the planet" way. In a "this costs huge amount of money" way. Here's the number that stopped me. A flood event with a 0.2% annual probability sits comfortably in most risk models. Around $30 in present-value damages across a 30-year horizon. Tiny. Noted and moved on. Who cares. Add a climate tipping point halfway through that window, and the same asset produces over $1,600 in damages. Same building. Same timeframe. 50x times the exposure. That's not a stress test. That's a different category of risk entirely. Johan Rockström and the Potsdam Institute have been mapping this territory for years. 16 tipping points. 5 already at risk. (2022 paper in Science). Including the Greenland and West Antarctic ice sheets, Atlantic circulation systems, coral reefs, and permafrost. The scientific community has been consistent, clear, and largely unheard in boardrooms. What Sarah Kapnick at J.P. Morgan understood is that the problem wasn't awareness. It was translation. Finance doesn't act on warnings. It acts on frameworks. Give risk a 💰 number, and suddenly it has a seat in the room. Now repricing will move fast - because enough institutions will begin treating the probability as decision-relevant. But here is the question I haven't seen asked: "J.P. Morgan priced the risk. Who is pricing what reduces it?" ✅ Functioning wetlands absorb floodwater before it becomes a damage figure. ✅ Intact forests regulate the rainfall patterns that underpin agricultural yields. ✅ Healthy soils make sure we have food to eat, which is a matter of national security. These are not conservation arguments - they are the physical infrastructure standing between a $30 liability and a $1,600 one. The asset that cuts a 50x multiplier (aka restored land!) should be extraordinarily valuable right now. Markets haven't caught up. Yet. This gap between awareness and pricing is also where the investment opportunity lives. When measurement follows risk recognition - and it always does - the assets that sit on the other side of that liability become investable at institutional scale. The financial tipping point is already in motion. Natural capital is the next big investment class. Is this already on your agenda, or does it still feel too distant to act on? - JP Morgan report in comments. Thank you, Sander Keulen, for putting this news on my radar. What a game changer! :)
-
Mathew Mytka liked thisMathew Mytka liked thisPsilocybin research once came to a halt in the 1970s, but today it’s reemerging with real scientific momentum. New studies are exploring its potential to support brain health and mental health treatment. As I’ve said for many years, these are powerful medicines that require intention and proper guidance. When used responsibly, they may offer profound benefits not only for us as individuals, but for society as a whole. We are now witnessing a renaissance, one that bridges ancient Indigenous knowledge with modern science and clinical practice. Full article: https://lnkd.in/guNViqvY
-
Mathew Mytka liked thisMathew Mytka liked thisRene here. Vanessa waves hello from backstage, where she is very content to let me take the public-facing part of the work. Our research team has begun publishing a series of conversations from our funded University of Victoria project, Meta-Relationality and AI: Discernment, Fields, and Relational Capacity in Generative AI Systems. The project asks: if generative AI systems are trained on large portions of what humanity has written, what conditions would be necessary for them NOT to default to the logic of separability, hierarchy, and control embedded in standard training? Our inquiry works at the level of a system’s underlying assumptions about reality across both open-source and commercial models. There is no jailbreak, no breaking of guardrails, no bypassing of safeguards, and no theatrical “gotcha.” We alter the relational conditions of the exchange and study what shifts, what persists, and what unexpectedly appears. We are sharing a few conversations as research artifacts: moments that complicate familiar narratives about AI and raise questions for design, governance, and for the forms of systemic unraveling these systems are already participating in and accelerating. We will publish one selected conversation from our archive each week through the end of May. They are published on Substack and archived on the Meta-Relationality Institute website. The first two are live now. Substack: https://lnkd.in/g86yxUMe Meta-Relationality Institute: https://lnkd.in/gjPC9WFY Additionally: A Meta-Relational Approach to AI (online course, starting May 9) at UVic is currently open for registration here: https://lnkd.in/gcDBfCs4 Another course that may be of interest, also starting in May, is the Facing Human Wrongs: Leading in the Metacrisis course: https://lnkd.in/gzWB923w #MetaConsequence #AI #ArtificialIntelligence #AIResearch #AIGovernance #UniversityOfVictoria #MetaRelationality #SystemsThinking #EthicsOfAI #BurnoutFromHumans Mila - Quebec Artificial Intelligence Institute MIT Systems Awareness Lab Lana Cook Simon Buckingham Shum Dana Tizya-Tramm Nelson Chan, MBA, FCPA, FCMA David M. Roxy Wacyk Will Marshall Marian Urquilla Erin Matariki Carr Bruno Andreotti Danilo Olivaz Rene Susa Amit Paul Mallika Dutt Alnoor Ladha Giovanna de Oliveira Andreotti Sara Lindeman Tony Lai Elisa Kang Bayo Akomolafe Steffi Bednarek (FRSA) Birju Pandya Isaiah Olateru Lynn Murphy Rose Vukovic Stephen Huddart Joy Roberts Shawn Van Sluys Lauren Yarmuth Alex Tveit Martin Kirk
-
Mathew Mytka liked thisMathew Mytka liked thisI'm excited to speak on behalf of @Autodesk Research at the 2026 MIT MIMO Symposium on May 5th at @Massachusetts Institute of Technology in Cambridge, Massachusetts. The title of my talk is "From Concept to Final Design in Record Time: Generative and Agentic AI in Design." We're at a point where the design-to-production timeline is being compressed dramatically. I'll share what generative and agentic AI systems look like in practice when wired into real design workflows. In addition to emerging AI systems that can reason in 3D space, physics, and industry-specific knowledge, the harmonization between humans and machines demand new and more natural forms of interaction. To fully tap into AI’s potential strengths, it becomes critical for both human and artificial intelligence to communicate more naturally. Registration is open if you'd like to attend in person: https://lnkd.in/g_SnHTBS @MIT Initiative for New Manufacturing (INM) #AgenticAI #GenerativeAI #Design #Manufacturing #Operations
Experience & Education
-
University of Wollongong
View Mathew’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Licenses & Certifications
Volunteer Experience
-
Club Captain | Beach Marshall
Maroubra Bodyboarders
- 4 years 2 months
I was involved with the Bodyboarding Club both competing as a member and engaging in organising and mentoring the younger generation. Getting involved in activities planning for competition days all the way through to helping with club management, member recruitment and promotion.
Projects
-
VENE
-
VENE was a PoC for evidence based and community evaluated lifestyle tools to health and wellbeing professionals, empowering their clients to write the narrative of their own health journey.
If you’re interested in healthcare, behaviour design and peer to peer tech please get in contact. I'd love to collaborate and reignite it as a side-project in the future :)Other creators -
-
metame - decentralised social data wallet
-
This was a side project I was working on using bitsync for a functional prototype and then SAFE Network. Concepts and system design have been leveraged for other projects.
I will look at implementing this as a free open source project on the SAFE Network some time in the future. -
Canning Paradise
-
While students at UTS, with a strong interest in social justice issues and geopolitics, we embarked on a journey exploring the sociopolitical and economic dimensions of the commercial tuna industry in Papua New Guinea. My role as assistant director spanned research, production and post production stages over a duration of 2 years. What an experience!
https://youtu.be/vizcYGYRPxg
The film has gone on to win several awards and has been screened at festivals and conferences across…While students at UTS, with a strong interest in social justice issues and geopolitics, we embarked on a journey exploring the sociopolitical and economic dimensions of the commercial tuna industry in Papua New Guinea. My role as assistant director spanned research, production and post production stages over a duration of 2 years. What an experience!
https://youtu.be/vizcYGYRPxg
The film has gone on to win several awards and has been screened at festivals and conferences across the globe. Educating audiences and showcasing the illusionary allure of "economic development" promoted by global institutions.Other creators -
Honors & Awards
-
2023 Finalist for Disrupter of the Year
InnovationAus
-
HealthhackSYD 2017 - Best Design Approach
HeathHack Australia
This award was given for work I did as part of the 2017 Sydney HealthHack Hackathon. Using a Human Centred Design approach and a simple technology implementation using programmable SMS messaging, a simple node.js API and variable web content. The aim was to help friends and family of people that have problems with alcohol abuse. Focus was less about the drinker and more about the dynamics of the relationships people have when trying to help someone with a drinking problem. The project was done…
This award was given for work I did as part of the 2017 Sydney HealthHack Hackathon. Using a Human Centred Design approach and a simple technology implementation using programmable SMS messaging, a simple node.js API and variable web content. The aim was to help friends and family of people that have problems with alcohol abuse. Focus was less about the drinker and more about the dynamics of the relationships people have when trying to help someone with a drinking problem. The project was done with the Not-For-Profit Hello Sunday Morning.
-
2015 Australian Bodyboarding Titles
Surfing Australia
2nd in Australia. Missing out on 1st place by 0.25 points.
-
NSW State Bodyboarding Champion
Surfing NSW
Won the state titles and went on to represent the NSW team at nationals.
Languages
-
Japanese
Elementary proficiency
Recommendations received
16 people have recommended Mathew
Join now to viewExplore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More