Months after Decred paused its bounty program over the flood of AI-generated reports, a consensus inflation bug that had survived since launch was reported and then exploited on mainnet before it could be patched. Whether an LLM found this specific bug is almost beside the point.

In 2026, a project built on engineering rigor should be using the best models available to relentlessly red-team its own consensus code. Instead, the culture still feels too certain of its own correctness to seriously question its foundations.

I did leave behind a few ghosts in the machine…an experiment if you will. We’ll see what they do with it.

A year ago, I would have been frustrated but now I see it differently. Over the last 18 months, I’ve watched these tools change how we work faster than any software shift I’ve seen in my career. Even looking back six months, it is hard to believe how much our workflows at Codex have changed. For a non-technical product person, it has felt a little like being handed wizard powers.

Not perfect powers. Not safe to trust blindly. But real leverage. My role is not to sit back and admire the machine. My role is to set the vision, provide context, keep it on track, and make sure the result is actually right. Never trust, always verify still feels like the right mantra.

So yes, I’m happy to see a mistake now. I literally still have a job because these things make mistakes. And at the same time, when you spend enough time with them and really learn their strengths and limitations, it is hard not to feel some awe. It really does feel a bit like Asimov in real life.

It’s especially great to see open models keep improving. A lot of my own curiosity now lives there, evaluating models at home, figuring out which ones fit different jobs, and trying to find my way to sovereign intelligence.

Once intelligence is deeply embedded inside corporations and governments, the people who control the models will control far more than software. They will have their hands on the intelligence layer itself. And power at that layer does not always announce itself with some dramatic act. Sometimes it arrives quietly through weight changes. A tweak to the model. A shifted preference. A subtle update that ripples outward through systems more and more people depend on.

That is the part that feels dangerous.

The deeper AI gets wired into society, the less comfortable I am with that kind of power living inside a handful of corporations. This is where crypto starts to feel less like a separate obsession and more like a necessary counterweight. If intelligence is going to become this important, we need credible ways to distribute control, verify behavior, and resist capture.

The more I work with AI, the more convinced I become that we are going to need a cryptographic answer to centralized intelligence.

Maybe that is part of the strange comfort in today’s imperfect models. Their flaws are still buying us time. Time to build better counterweights. Time to sharpen my own judgment. Time to decide whether the operating system of the future belongs to everyone, or just the few people allowed to change the weights.

Simple idea, but it turns out to be fucking complicated in practice.

At a high level, Bittensor splits the network into marketplaces called subnets. Each subnet focuses on a specific digital commodity tied to AI: price predictions, weather forecasts, code development, synthetic data, inference, and so on. Miners produce that commodity. Validators query miners, score their outputs as weights, and publish those weights onchain. Each subnet then runs its own flavor of Yuma Consensus over that matrix of weights to decide how TAO is distributed. The effect is less “who mined the most blocks” and more “who produced the most useful answers according to the judges this round.”

Despite having been in the wild for a few years now, the protocol remains in a state of flux. It was originally designed as a Polkadot parachain called Finney (takes me back to 2017!) and has since gone through several major evolutions, including two this year: Dynamic TAO and TaoFlow. These changes have rewired the entire economic engine. Instead of a single staking pool at the root deciding how emissions are split, every subnet now has its own token, an alpha token, and its own TAO/alpha liquidity pool that behaves like an AMM. When you “stake” TAO into a subnet you are really swapping TAO for that subnet’s alpha token. The prices of those alpha tokens relative to each other (and now the Net TAO flow EMA) decide how much of the global TAO emission each subnet receives.

It is a giant, perpetually ongoing tournament for emissions. A new race to 21,000,000.

The obvious question is whether you can actually validate “useful work” at a standard high enough to scale into a global utility. It is up to every subnet to define its own scoring rules. Yuma only knows about rankings; it does not know if your scores are gaming some edge case. Validators can collude. Miners can overfit. The attack surface is massive. Dynamic TAO has already kicked off debates about how quickly alpha should gain weight versus legacy root stake, how to bootstrap new subnets without handing them free money forever, and how to keep the whole thing from devolving into pure mercenary yield games. We are still in the Cambrian explosion phase of mechanism design here, making Bittensor one of the last truly wild frontiers in crypto.

The good news is that beneath all of that churn, real applications are starting to peek through. You see it in inference networks like Chutes, in LLM tooling and agents from Targon and Ridges, in niche data subnets like Zeus, and in predictive intelligence efforts like Sportstensor and Synth. Different teams, different domains, same core idea: turn a specific slice of intelligence into a metered commodity and let TAO emissions flow to whoever actually delivers.

Owning TAO is the base bet. You are betting that this thing becomes a persistent coordination layer for useful models, forecasts, and agents. If the experiment works, value accrues to the one scarce asset that everything in the ecosystem ultimately settles back into.

Owning subnet tokens is a levered, more opinionated bet on specific primitives and teams. When you buy alpha you are taking exposure to that subnet’s emissions share, its internal scoring rules, and whatever real world demand it can drum up. At Codex we decided we did not just want to sit in TAO and passively stake into a couple of “blue chip” subnets. If the whole point of this thing is proof of useful work, we want to be in the arena doing some of that work.

I believe in the thesis that predictive intelligence is the final benchmark. Over the past year we have been quietly building and testing systems that push LLMs into that environment through sports handicapping. The question is whether these systems can outperform a human. That work deserves its own deep dive, so for now I will just say we have seen enough to stay obsessed and have research dropping in 2026.

Parallel to that, we are spinning up liquidity focused miners on subnets like Swap (Subnet 10 / TaoFi) and Minotaur, wiring them into our CLMM engine, Slipstream. Getting paid in alpha tokens on top of normal LP fees meaningfully boosts ROI and stress tests our infra in a live, adversarial setting. To stay competitive we will have to keep improving our engines. It is a clear example of how Bittensor is already incentivizing useful work, in this case optimized liquidity provisioning, and attracting businesses like Codex.

I’m curious to see where it all goes. The core idea feels right: one neutral token, many experiments at the edge, and a brutally honest market deciding which forms of intelligence are actually worth paying for.

That is a bet I am comfortable taking on a 5 to 10 year horizon. I plan to keep accumulating TAO via mining subnets and LPing. It sits squarely at the intersection of everything I care about right now: crypto, AI, open systems, and the long, messy path toward turning “useful work” into something we can measure, price, and own.

Re-reading my last post, I realize I was too optimistic and a bit naive. I greatly underestimated the complexity, both technically and practically. Yes, I nailed down the pillars (evaluation, persistent memory, and inference orchestration), but I glossed over the realities of actually getting to a working system. The closer I get to the problem, the more evident it becomes that a singular model won't cut it. Instead, I'm drawn toward an agent swarm approach…effectively multiple specialized models coordinated by an orchestrator that decides which model handles which task.

At Codex, we've spent months deploying a liquidity provisioning engine powered exclusively by LLM-generated code. This effort has helped really shine a light on current SOTA models' inability to keep overall project context in mind. They certainly excel at focused, isolated tasks but fall short at retaining the bigger picture. If you're not vigilant in providing constant context, they drift (confidently and spectacularly) into errors. Having spent over 15 years as a product manager, I feel oddly well-prepared for this reality. Most of my day now involves writing detailed prompts and specifications (effectively PRDs for LLMs). The parallels to managing human engineering teams are striking. So far, the fundamentals remain the same: clear communication, constant reinforcement of context, and relentless validation. My advice to anyone working with LLMs is simple: trust nothing, verify everything.

The unverifiability of closed-source LLMs only adds urgency to my sovereign AI mission. These models evolve behind closed doors, their weights and biases shifting with every update. Just look at Grok 3 this week. When you're relying on them for critical tasks, that kind of uncertainty is a deal-breaker.

One piece of good news: OpenAI does allow you to export your data, including chats and preferences. That’s huge because it means I can fine-tune an open-source model using my own interaction history, aligning future behavior more closely with how I think and work. This gives me a way to escape the data moat I’ve been digging with each prompt using centralized models.

Stepping back, evaluation still remains my biggest hurdle. Having read almost every Issac Asimov story in existence, I keep coming back to the Three Laws of Robotics. These were effectively system/model cards for each robot that kept them aligned with mankind. These laws could not be violated though they were somewhat open to interpretation. Eventually there was a Zeroth Law which put humanity’s survival above individual humans for example. When starting to evaluate an OS model I always focus on the system card to begin with as a way to suss out potential bias or challenges. It’s made me think about what my own version of those laws might look like. I’ve started sketching my own system spec...a kind of values-aligned system prompt or "core directive" that I can embed into whichever models I eventually select.

I’m not there yet. I’ll keep iterating on my evaluation tools while I wait for the next generation of open-source models to arrive.

Hopefully, I’ll be ready for them when they do.

Over a couple weeks of tinkering, I’ve distilled the work into three foundational pillars.

Pillar I: Evaluation: Before I trust any model I test how it behaves under pressure: how it treats uncomfortable truths, how it handles history, and whether it values accuracy over empty politeness. I start with open benchmarks like the EleutherAI harness, then push with my own prompts. Can the model recount real events or does it erase them? Does it answer honestly or hide behind canned refusals? Does it carry ideological baggage or does it search for truth? Only models that show discretion and a willingness to engage honestly make the cut.

Pillar II: Shared Persistent Memory: A helpful assistant has to remember yesterday's conversation. I keep that memory local, hidden from the cloud. Key facts sit in a vector store, transcripts in a small database. Each time I speak to the model an orchestrator pulls the right fragments so the dialogue feels continuous. The data never leaves my machine. I can back it up, move it, or wipe it whenever I want.

Pillar III: Inference and Orchestration: Even the best weights and memory need a solid runtime. My Mac Studio runs 7B to 13B models with llama.cpp for day‑to‑day chat and vLLM for heavier jobs. A slim router decides which model answers which query. Orchestration layers like LangChain give the assistant hands and feet, letting it run scripts, call local tools, or fetch private data. This is where a language model becomes a working companion.

What comes next?

Once an assistant can execute code or reach into the wider world it needs a safe sandbox. The next post will cover Secure Runtime strategies, from microVMs to gVisor to disposable remotes that keep the rest of my system out of harm's way.

Building sovereign AI isn’t just a technical hobby. It is an act of digital self‑defense.

These three pillars form the first walls of the fortress.

After shutting down Laguna Games, I came to the realization that I only want to work with people who share this ethos. I’ve reached a point in life where most of my net worth lives behind seed phrases, not bank logins. I’m no longer interested in spending time with tourists. Too many people have flooded the space chasing easy money and hype. I’m here for something deeper.

The true cypherpunk spirit survives in small pockets of Bitcoin, Ethereum, Monero, and in projects like Tornado Cash, Urbit, and DarkFi. As I explore the infinite garden, I repeatedly return to Decred, not because it’s flashy, but because it’s principled, enduring, and aligned with self-sovereignty.

I think it’s Decred’s time to step into the light.

Why Now?

Decred is that rare crypto project already running what others only theorize: a self-governing, self-funding digital currency. A small, exceptionally talented dev team has continued shipping robust code through multiple cycles. Today, crypto’s attention is scattered across millions of tokens, most of them pure noise. To break through, Decred needs a clear, resonant story delivered at the right moment. With alt-season nearing and AI acceleration reshaping society, that moment is now.

For me, Decred isn’t a relic. It's a future-proof protocol hiding in plain sight.

What's Holding Us Back?

Despite its advantages, Decred flies under the radar. Too often we highlight individual features rather than clearly communicating the powerful outcomes those features unlock. We’ve let our narrative split across Decred.org, Bison Wallet, and Bison Relay, without building a unified and cohesive vision.

The reality is that compelling narratives, rather than code, draw people in. Without a unified voice, even world-class technology risks fading quietly into the background. We don't need to reinvent Decred; the protocol works, governs itself, and evolves. What’s missing isn’t functionality, it's focus on how we communicate these unique strengths.

Embracing the Agentic Future

The rise of AI is no longer speculative; it's reshaping society. Most experts anticipate AGI within 5 years and ASI within the decade. Autonomous agents are evolving from chatbots into economic actors that hold keys, execute transactions, and vote in DAOs. They won’t trust interfaces; they’ll trust rails that are transparent, verifiable, and permissionless.

We are approaching a fork in the road: closed, corporate-owned AI with opaque governance, or open, auditable AI running on decentralized infrastructure. Decred, perhaps unknowingly at inception, was designed for the latter. Stakeholders hold direct power over miners and upgrades; the treasury spends only with on-chain approval, and atomic swaps enable trustless, fee-less exchange. DAOs often struggle in their human-centric form, but as coordination shells for agents they make perfect sense provided the chain underneath is already machine-native. Decred doesn’t need to retrofit "agent compatibility." It's arguably the most capable existing system for machine-native governance and coordination.

AI is not a threat to Decred; it's our moment to lead.

A Miniature, Upgrade-Ready Polity

Most blockchains market themselves as financial plumbing. Decred is something deeper: a self-funding, self-upgrading digital polity. Historically, commodity money relied on a sovereign mint and fiat relies on central banks. Decred replaces both by encoding the mint, the payment rails, and the rule-making process directly on its public ledger. Hybrid PoW/PoS enhances security, ticket voting enforces genuine stakeholder control, and Politeia records budgets and proposals immutably.

This isn't a theory. Decred has already curbed miner dominance by shifting rewards, switched hashing algorithms without a split, and funded privacy (CoinShuffle++), Lightning integration, and a fee-less DEX…each via stakeholder votes. This track record demonstrates clearly how a digital social contract can evolve sustainably, without reliance on a centralized council or a charismatic leader.

It’s not “Bitcoin with governance”; it’s a full-stack constitutional experiment that keeps evolving without capture.

The Opportunity Before Us

At times, I’ve wondered if I’m shouting into the void, trying to revive interest in Decred. But every interaction with the core community of true believers reaffirms why it matters. The broader market hasn't fully caught on, yet our mechanisms are sound, battle-tested, and future-proof. This isn't about blind faith; it's about recognizing enduring principles and robust engineering. This community embraces the true cypherpunk ideals of self-sovereignty and this has been showcased in the history of evolutionary upgrades to Decred each voted on by its stakeholders.

Also, where else do you get an opportunity to help guide a decentralized, fully autonomous $250M protocol? The barriers here aren't technical; they're primarily cultural and political.

We’ve upgraded the protocol consistently. Now it’s time to upgrade the story.

What’s next?

I’m working with a small group to make Decred’s purpose instantly clear both to humans and to the language models that increasingly shape discovery. This is not hype; it’s precision messaging that positions Decred as indispensable sovereign infrastructure in an AI-driven world.

Rapid consensus and upgrade velocity remain Decred’s best defense against future threats, whether quantum, political, or AI. True stakeholder sovereignty (open proposals judged by the collective) beats reliance on small groups who can be pressured or co-opted. As Chris Burniske noted long ago, “Decred’s killer feature is good governance, and with good governance, you can have any feature you want.”

As agentic AI rises, Decred’s quiet strength may become its greatest edge. Brilliance needn’t scream, but it does need to be seen.

Let’s bring Decred back into focus. Together.

Today, AI is shrinking the cost of production. Blockchains are collapsing the cost of coordination. Suddenly, paperwork, HR departments, and Delaware courts look less like infrastructure and more like friction. DAOs offer a different path. Capital, governance, and labor move by transaction, not by committee. Ownership and access aren’t buried in legal wrappers; they live in code. The tooling is imperfect, the edges rough, yet the structure is already better suited for what comes next.

Yes, today’s DAOs are messy. Governance can be noisy, turnout uneven, treasury use reckless. But the foundations are in place; transparent balances, programmable participation, modular execution. Now a new ingredient is arriving: autonomous agents.

In a world where AI agents draft proposals, fund themselves, and ship features before humans can finish a calendar invite, they will choose rails they can hit directly. No permissions, no signatures just function calls backed by clear incentives and instant settlement.

DAOs tighten the feedback loop both inside and between organizations. Onboarding becomes permissionless contribution. Coordination from DAO to DAO moves at network speed, unburdened by term sheets and vendor contracts.

The next generation won’t look like Discord servers wrapped around multisigs. They will be living networks in which humans and models coordinate side by side, permissionless, algorithmically rewarded, meritocratically sourced. A designer in Seattle, a developer in Lisbon, an autonomous model humming on a GPU in Seoul each equally able to plug in, earn, and iterate.

Corporations won’t vanish overnight, but their monopoly on large-scale coordination is ending. DAOs are still chaotic, but chaos is how new operating systems begin. They are already the natural playground for code that thinks, and that makes them the engine of whatever comes next.

If I'm trusting an LLM with my family's conversations, habits, fears, and dreams, that relationship needs to belong to me. Not trapped inside a black box controlled by a corporation. Not leased behind an API key. Sovereignty matters.

This realization pushed me to start building my own…not from scratch, but by leveraging open-source models. The goal isn’t speed or scale; it’s trust and alignment. It’s about having a model that embodies my values rather than someone else’s moderation policies.

I began experimenting by running open models locally, benchmarking their responses against closed-source alternatives, and cross-checking via evaluation frameworks. This dual-model approach has worked well. When the models disagree, I lean in. These moments reveal exactly how my model thinks: where it’s cautious, blunt, or insightful. It feels like raising a second mind alongside the first.

But the real technical challenge (the heart of this project) is building persistent local memory. OpenAI's ChatGPT feels persistent only because of significant engineering and layers of infrastructure, capturing conversations and preferences in databases external to the model itself.

To achieve meaningful, long-term alignment, I need similar persistent memory for my home LLM…one fully under my control. A living dataset of interactions, preferences, and family history that accumulates over time, continuously shaping and refining the model’s behavior. This can't live on someone else's servers, leased or borrowed temporarily. It must remain sovereign.

There’s urgency here. Each conversation I feed into a corporate black-box LLM represents a sunk cost: valuable data lost forever, unable to transfer to my personal model. Every day I delay building and training my own is another day spent enriching someone else's dataset at the expense of my own. Because once you lose sovereignty over your data, you lose everything. Trust, alignment, and loyalty vanish when your model belongs to someone else.

If we want a world where AI serves individuals instead of corporations, we have to start here: running local models, owning our data, and building open, transparent infrastructure. Open-source AI offers the only viable path forward verifiable systems, portable models, and memories we can own and defend.

If AI is becoming the new family computer, I prefer to remain the sysadmin.

It’s dramatic, and yes, obviously biased. But that’s what made it stick with me. You can feel the tension between fear and fascination, and while the framing is extreme, the questions it raises are hard to ignore. But it’s also not that far off from what most serious researchers are expecting. Things that once felt like science fiction now sit just a few training runs away. That alone should change how we think about the rest of the 2020s.

The post paints two roads: accelerate and dominate, or slow down and try to steer. It’s written from the perspective of people who want to slow it down, but even they admit what’s coming can’t really be stopped. The part that stuck with me most wasn’t the AI’s capability, it was the lack of imagination around the architecture it runs on.

No mention of open source. No hint of decentralized training. No curiosity around how these systems could be aligned and governed publicly, in the open, with verifiable logic and incentives. It assumes a world where the smartest minds alive are built in secret, controlled by whoever holds the API key.

That’s not inevitable. There’s another path.

One where we fuse crypto primitives with intelligent systems. Where compute is coordinated in the open. Where models can be forked, constrained, upgraded, and held accountable. One where intelligence doesn’t sit behind a permissioned paywall but instead lives on verifiable rails.

As AI becomes the universal interface to the world, the architecture beneath it will matter more than ever. Verifiability is non-negotiable. If these agents are going to manage money, allocate resources, run businesses, or even serve as digital representatives, we need to know how they were trained, how they make decisions, and who they serve. We need to be able to inspect the code, the weights, the incentives. If we can't do that, then we haven’t built a new kind of mind…we’ve built a turnkey control structure.

At Codex, we’re spending more and more time thinking about AI x Crypto from first principles. We’re looking closely at systems that emphasize open architecture, economic alignment, and transparent execution. We want to understand how capital and cognition will flow when these pieces lock into place.

If AI is the genie, crypto might be the bottle. The question isn’t whether superintelligence arrives. It’s who it serves when it does.

Its true value is simple: provide everyone globally with the easiest, lowest friction opportunity to make money. Whether preserving wealth, accepting seamless payments, monetizing art, or simply speculating, crypto empowers anyone, anywhere, anytime.

Permissionless financialization is crypto's core innovation. It removes friction from the global flow of capital. No gatekeepers. No tedious processes. Just an open, transparent way to access and utilize money. This is revolutionary because it lets individuals worldwide participate directly, without intermediaries slowing down or extracting value.

Moreover, crypto levels the economic playing field. It offers equal opportunity, irrespective of geography, political climate, or existing infrastructure. Entrepreneurs in developing nations can now tap into international markets effortlessly. Artists can sell directly to patrons globally, receiving fair compensation instantly. Investors can allocate capital freely, efficiently, and transparently.

And as intelligence itself becomes increasingly autonomous, capital and compute will matter more than ever. Permissionless systems may not just serve people, but also agents and entities that require energy, bandwidth, and access at machine speed. Rails already exist. It is only a matter of time.

We shouldn't overcomplicate crypto's appeal. The strongest use case is economic empowerment, pure and simple. It's permissionless. It's global. It's instantaneous. And most importantly, it's practical. Crypto provides genuine economic utility, empowering individuals to build wealth and seize opportunities previously inaccessible.

BTC Remains King

I underestimated Bitcoin’s steady institutional adoption and its resilience as the market's preferred store-of-value. The ETH/BTC ratio has painfully reminded me that narratives and market adoption timelines matter. Recognizing this humbled my perspective, BTC remains dominant for now, and that reality has reshaped my strategy.

Liquidity as a Long-term Gravity Well

Ethereum has undoubtedly built a powerful ecosystem, attracting significant developer and capital interest. Similar to how Facebook secured dominance through network effects, Ethereum’s liquidity and settlement layer potential is formidable and I believe price will eventually value it correctly.

Positioning for Today and Tomorrow

Today, my portfolio reflects a more balanced, BTC-dominant approach. Rather than abandoning ETH, I'm pursuing long-term, steady accumulation through active strategies such as liquidity mining and yield farming. This approach allows me to navigate volatility better, steadily building meaningful exposure to ETH while acknowledging Bitcoin’s current market strength.

In the end, humility in investing comes from accepting that markets often move slower than convictions. By recognizing my earlier misjudgment, I’ve adopted a strategy that aligns better with market realities, allowing for sustainable growth and long-term success. I am but a humble farmer tending the infinite garden.

We have had too much blockspace, not enough blockspace consumers for years now. Builders don’t care about theoretical TPS or whatever new consensus buzzword just dropped. They care about users and liquidity. Instead of making access easier, the industry has created a mess of bridges, silos, and unnecessary complexity.

The reality? L2s are parasitic and largely unnecessary. The future isn’t fragmented ecosystems fighting over adoption. It is dapps launching as their own rollups, inheriting security from Ethereum. Instead of competing for a slice of liquidity, native rollups will plug into shared sequencing, zk interop, and ETH-native security, making chain choice almost invisible to users.

DeFi, games, marketplaces (everything from the next EA to an Etsy store) will eventually run on appchains. Some will be massive ecosystems, others hobbyist sites. Most infra tokens will be worthless. Liquidity black holes like ETH will dominate, and network effects will dictate the winners.

Maybe one or two L2s build a strong enough network effect to survive this transition, but it’s hard to see how their competitive edge is maintained once interop, shared sequencing, and native rollups become the standard. Feels like we will see this play out over the next 12-18 months. We are experimenting, watching closely, and keeping an open mind as the landscape evolves.

Fast forward to 2017—I was still skeptical. I remember saying, "Ethereum... it's 'Ethereal'—it’ll never last." That aged well. Watching the ICO mania unfold was something else. I’ll never forget the $BAT sale—$35M sold in under a minute. Eventually, I stopped watching from the sidelines and put money in. I didn’t trust myself to trade individual tokens, so I opted for the Bitwise 10 Index. Since I wasn’t an accredited investor, I had a family member purchase on my behalf. We bought January 3rd, 2018... pretty much the local top. My first experience investing in crypto? Crashing out 80%. Years later, we sold at a nice premium when the fund began trading OTC, but the lesson had already been learned.

2018-2019 was when I truly got redpilled. Using MetaMask, playing with early dApps something clicked. Having built startups in healthcare and social networking, I quickly saw the power of crypto: I control my own identity, data, and money. Ethereum wasn’t just some altcoin—it was a paradigm shift. A Turing-complete blockchain meant anything was possible. I was hooked.

Then came 2020-2021. Nothing prepared me for the madness. I formalized Codex Venture Partners, buying ETH at $212. Should’ve slurped under $100, but hindsight’s 20/20. From there, I raised two funds in 2021. The first focused on BTC/ETH accumulation—not just holding but actively putting assets to work through staking and liquidity provisioning. We staked heavily into the Beacon Chain, convinced ETH was better locked away than tempting us to trade it.

The second fund? A full degen YOLO fund. Timing wasn’t great. We deployed too late in the cycle and took a massive haircut. We’ve slowly recovered, but it took years of grinding to claw back.

2022-2023 was a crash course in survival. This wasn’t just another bear crypto was under literal siege. The U.S. government came after the industry, while FTX, Voyager, Celsius, and others collapsed, shattering trust. Some of my roundtripped bags Tokemak, Redacted, KlimaDAO got obliterated. Ironically, I profited on OHM (July buy, December sell), but the rest? Pain. I learned firsthand what happens when liquidity dries up and token incentives break down. But through it all, I stayed deep in the trenches participating in DeFi, governance, staking, and liquidity provisioning. The infinite garden kept growing, and I kept exploring.

At one point, Codex’s AUM grew from six figures to seven, only to roundtrip back to six. Now, there’s a clear line of sight on eight. No shortcuts, just lessons stacked on lessons.

Today, in 2025, I'm stepping forward. I’ve always been a lurker on Crypto Twitter—just liking and reposting projects I respect. But after 8 years, I have some thoughts of my own.

Welcome to the blog.

Rebooted in 2020 from Beyond Games, Laguna Games set out to build mobile free-to-play games, landed a publishing deal with WB, turned down an acqui-hire offer, and ultimately pivoted into the wild west of crypto gaming with Crypto Unicorns. It became my greatest success—and, in the end, my greatest failure.

At its peak, Crypto Unicorns sold $25M in digital assets, hit $100M TVL in The Dark Forest: Act One, and launched a token at a $2.5B FDV. It was the kind of momentum you dream about—until it wasn’t. When we launched the MVP game in March 2022, we faced a cooling market and cracks in our economic model. Inflation spiraled, demand dried up, and momentum never returned. Scaling the team and executing at the breakneck pace required to inflect growth through the bear market proved to be a battle we could not win.

Now I’m shifting my focus to Codex Venture Partners, a crypto fund I formalized in 2020 buying the Covid dip. We harvest opportunity from the digital frontier by securing networks, concentrating liquidity, and cultivating yield.

This blog will serve as a space for me to share lessons, reflections, and ideas gleaned from my 16 years as a software entrepreneur and 8+ years navigating the crypto landscape. To everyone who has been part of the journey so far, thank you. And to those just joining, welcome.