Alpin Yukseloglu: AI will revolutionize crypto security, superhuman auditors are on the horizon, and emerging markets offer high-yield opportunities | Bankless

Key takeaways

• AI is poised to significantly enhance security in the crypto industry, raising the industry’s potential.
• Improvements in AI models for detecting smart contract vulnerabilities are progressing rapidly.
• Superhuman AI auditors may emerge soon, challenging current security assumptions in crypto.
• The crypto industry is already hardened against intelligent adversarial actors.
• Emerging markets offer lucrative investment opportunities with high yields.
• Bricks bridges DeFi with traditional finance, enabling access to real collateral and structured products.
• The perception of technology as a threat can lead to unnecessary security paranoia.
• Superintelligent AI’s impact on security dynamics is uncertain, favoring neither offense nor defense.
• Fundamental constraints exist that even superintelligence cannot overcome.
• Acceptance and denial both imply a lack of control over future outcomes.
• Technology’s future is best understood through experimentation rather than predictions.
• Engaging with technology actively can mitigate fears of new advancements.

Guest intro

Alpin Yukseloglu is an Investment and Research Partner at Paradigm, a research-driven crypto venture firm with over $12.7 billion in assets under management. Previously, he served as a protocol engineer and product lead at Osmosis, bringing deep technical expertise in blockchain systems. He co-authored the EVMbench benchmark with OpenAI, an evaluation framework that measures how AI agents detect, patch, and exploit smart contract vulnerabilities—work that revealed AI’s capability to identify over 70% of critical fund-draining bugs.

AI’s impact on crypto security

• “AI will significantly enhance security in the crypto industry over time.” – Alpin Yukseloglu
• “In the long term it’s now increasingly clear that AI is going to be extremely good for crypto.” – Alpin Yukseloglu
• The improvement in AI models for detecting smart contract vulnerabilities is rapid and significant.
• “When we started working on evm bench… the models were able to find less than 20% of the bugs… this number went up to over 50%… it jumped up to over 70%.” – Alpin Yukseloglu
• Superhuman AI auditors may emerge by the end of the year, challenging current security assumptions.
• “I’m pretty confident at this point by the end of the year a superhuman AI auditor… will just completely break all of our assumptions.” – Alpin Yukseloglu
• Crypto has been operating under the threat of highly intelligent adversarial actors, making it relatively hardened against attacks.
• “Crypto is already quite hardened… we already have existed in crypto under this threat model of extremely intelligent adversarial actors.” – Alpin Yukseloglu

Emerging markets and investment opportunities

• Emerging markets generated over $115 billion in annual yield for investors, with yields ranging from 10% to 40%.
• “In 2024 emerging markets generated over a $115,000,000,000 in annual yield for investors.” – Alpin Yukseloglu
• Bricks connects DeFi with institutional-grade tokenization and compliance, allowing access to real collateral and structured products.
• “Bricks connects these worlds with institutional grade tokenization local banking rails compliance across jurisdictions.” – Alpin Yukseloglu
• The integration of DeFi with traditional finance is crucial for accessing real-world yield.
• “BRX does the heavy lifting so DeFi can finally access real collateral and structured products on top of real-world yield.” – Alpin Yukseloglu

Perceptions and realities of technological threats

• The perception that only bad actors will exploit technology leads to a psychosis around security threats.
• “I think the model of like there are only bad people in the world and they’re going to have access to this technology.” – Alpin Yukseloglu
• Superintelligent AI may not guarantee an advantage for either offense or defense in security.
• “Right now it’s not clear whether this is gonna be an offense or defense favoring technology.” – Alpin Yukseloglu
• There are fundamental constraints in the world that even superintelligence cannot overcome.
• “I will say that there are still fundamental constraints in the world like you can’t break laws of physics.” – Alpin Yukseloglu

Philosophical implications of technological advancement

• Both acceptance and denial imply a lack of control over future outcomes.
• “Peter Thiel has this framing where acceptance and denial most people relate to them as opposites.” – Alpin Yukseloglu
• The future of technology is best understood through experimentation rather than theoretical predictions.
• “The current frontier and like I guess you can argue the frontier has always been experimentally bound.” – Alpin Yukseloglu
• Agency can be developed through practical action rather than just faith.
• “I think so faith faith is good but it’s not a particularly agency inducing headspace to be in.” – Alpin Yukseloglu

Navigating uncertainty in the crypto space

• Exploring risks and integrating crypto into innovative labs can provide pathways to navigate uncertainty.
• “You can go figure out to what extent these things are at risk and then also start making headway into the labs.” – Alpin Yukseloglu
• The evolution of crypto has led to a clearer understanding of its use cases, particularly in stablecoins and prediction markets.
• “We have this store of value use case we have stablecoins that are compounding at this monstrous rate.” – Alpin Yukseloglu
• The intimidation of new technologies can be mitigated by actively engaging with them and taking agency.
• “The only reason why the singularity staring into the void is intimidating is because what all of these technologies are doing.” – Alpin Yukseloglu

Strategic approaches in a fast-paced environment

• In the current environment, moving fast and adapting is more valuable than taking time to plan.
• “The current environment we’re in because the frontier is so unknown and so unknowable to some extent.” – Alpin Yukseloglu
• Simple, long-established contracts are generally safer than newer, less tested ones.
• “Simple contracts that have been around for a long time I think are probably better in a better position.” – Alpin Yukseloglu

Vulnerabilities and economic dynamics in DeFi

• Smaller, less secure protocols are likely to be the first victims of exploits driven by AI.
• “I think there will probably be this canary in the coal mine effect where there will be smaller or protocols.” – Alpin Yukseloglu
• The cost of exploiting smaller contracts will lead to their disappearance as AI makes attacks more feasible.
• “When the cost to exploit a $1,000 contract is like you know 10 to $50 of tokens then those contracts just simply won’t exist.” – Alpin Yukseloglu

Enhancing AI capabilities with agent harnesses

• The agent harness enhances the capabilities of AI models by providing specialized tools for testing smart contracts.
• “The agent harness that we released is sort of not at the frontier of capabilities because we don’t want it to be used for black hats.” – Alpin Yukseloglu
• As AI improves, it will increasingly absorb the functionalities of the tools that support it.
• “Most of these tools that we add in fall like flake off a time because as the model gets better it just absorbs the harness.” – Alpin Yukseloglu

Advancements in AI model training and performance

• The harness acts as a bootloader that enhances the agent’s performance by providing an environment for testing.
• “The harness is kinda like a bootloader to get it started… it turns out that just giving an agent the ability.” – Alpin Yukseloglu
• The benchmark tool significantly reduces the false positive rate in bug detection to nearly zero.
• “We leaned on this to lower the false positive rate down to basically zero so it got to a point where if the agent tells you.” – Alpin Yukseloglu

Verifiability and model learning in crypto environments

• The verifiability of environments in crypto allows models to learn effectively and improve their performance.
• “The verifiability ended up being very important… the verifiable stuff is very easy for the models to learn.” – Alpin Yukseloglu
• AI models will become extremely proficient in analyzing crypto-related code very quickly.
• “I think the general trend and trajectory of you know these models are gonna get extremely good at crypto extremely fast.” – Alpin Yukseloglu

AI’s role in auditing and bug detection

• AI models are approaching the effectiveness of human auditors in finding critical bugs in smart contracts.
• “Chad GPT five point three codex is like 70% as good as all of the human auditors out there.” – Alpin Yukseloglu
• The AI models are capable of finding a diverse set of critical bugs, not just one type.
• “This is a very diverse set of bugs that it was able to find.” – Alpin Yukseloglu

Challenges and opportunities in crypto valuation

• Crypto has been stigmatized and remains illegible to AI labs, which has hindered its valuation.
• “The fact that there hasn’t already been a massive push for crypto related valuations is kind of absurd.” – Alpin Yukseloglu
• The slow adoption of crypto in AI labs is largely due to social factors and reputational volatility.
• “My sense is that it’s almost entirely a social thing… it’s very reputationally volatile.” – Alpin Yukseloglu

Skill disparity and reputational volatility in crypto

• The disparity in skill levels within the crypto industry can lead to a distorted perception of the sector.
• “The gap between the best people in the industry and the median person in the industry is much larger than anywhere else.” – Alpin Yukseloglu
• There is significant reputational volatility in crypto that can be advantageous for certain investors.
• “A lot of us have benefited from the fact that there’s significant reputational volatility.” – Alpin Yukseloglu

Bridging the gap between crypto and AI

• The lack of a strong brand bridging crypto and AI has hindered collaboration between the two fields.
• “There just hasn’t been a a brand that can bridge the crypto and the ai worlds.” – Alpin Yukseloglu
• Models can learn complex concepts in crypto with less direct training data due to the verifiability of the substrate.
• “There’s this dynamic where if you teach a model a poem in english and then biology in spanish.” – Alpin Yukseloglu

Opening the floodgates of crypto data

• The floodgates of crypto data for training models are starting to open, leading to new capabilities in AI.
• “Do you think that this right now is the time that the the floodgates of crypto data to train these models.” – Alpin Yukseloglu
• Security capabilities in crypto are expected to develop quickly due to the intelligence-bound nature of the technology.
• “I expect will develop very quickly… security capabilities… it’s extremely intelligence bound.” – Alpin Yukseloglu

Mechanism design and market innovation in crypto

• The implications of mechanism design in crypto markets represent open fertile soil for innovation.
• “I think for example things in the domain of mechanism design or around market related films… these are i think open fertile soil.” – Alpin Yukseloglu
• Current models are not very effective at executing complex on-chain transactions but are expected to improve rapidly.
• “These are all things that actually the models are not that good at right now but they will get good at really quickly.” – Alpin Yukseloglu

Long-term security improvements and asset management

• In the long term, improvements in security will positively impact the crypto industry by allowing more assets to securely remain on-chain.
• “You mentioned the in the long term crypto’s positively levered to almost all of these developments.” – Alpin Yukseloglu
• The future of crypto security is uncertain and depends on the industry’s proactive measures against potential threats.
• “We don’t know exactly who whether the attackers you know the black hats will get capabilities before the white hats do.” – Alpin Yukseloglu

Evolution of transaction speed and expressivity

• The evolution of transaction speed and expressivity in crypto addresses the double spend problem.
• “If you start from the first principle’s vantage point of let’s say you want to do payments at the speed of light.” – Alpin Yukseloglu
• The future financial system will likely consist of extremely secure digital assets and traditional physical assets.
• “It almost creates kind of a a barbell model of a security for for the world for financial assets.” – Alpin Yukseloglu

Crypto’s alignment with modern transaction needs

• The crypto industry is fundamentally aligned with the needs of agents operating at internet speed.
• “If you have agents that want to move at the speed of the internet and the current banking system was created before cars were invented.” – Alpin Yukseloglu
• There are strong network effects in crypto that will drive its adoption and success.
• “There are extremely strong network effects inside of crypto.” – Alpin Yukseloglu

The role of EVM in crypto development

• The Ethereum Virtual Machine (EVM) is currently the most common programming environment in crypto.
• “The evm is by far the most common programming language programming environment in in crypto.” – Alpin Yukseloglu
• Closed source contracts may provide a unique advantage for model training in crypto development.
• “If you take the world view of actually if it’s open source it gets in the training set.” – Alpin Yukseloglu

Formal verification and software security

• AI-based formal verification can enhance software security by quickly checking if components function as intended.
• “I think it is a real thing… formal verification is one way to quickly check whether a component of software is actually doing what it says it’s doing.” – Alpin Yukseloglu
• The future of software development will increasingly rely on formal verification due to the growing volume of software.
• “Part of their thesis… is that there’s more software that is being generated than can be possibly reviewed by humans.” – Alpin Yukseloglu

Reducing bugs with formal verification

• Formally verified software may have a lower surface for bugs compared to traditional coding.
• “You can make the case that actually the surface for bugs in writing a formal verification spec might be lower than writing the code to start with.” – Alpin Yukseloglu
• The best software will likely be formally verified in the future.
• “Definitely with time i think all of the best all the best models all of the best software will probably end up being formally verified.” – Alpin Yukseloglu

Crypto’s leverage to global developments

• Crypto is positively levered to developments in AI and other global changes.
• “I think that you know we talked about how crypto is positively levered to the security developments in ai.” – Alpin Yukseloglu
• The technology behind crypto is on a compounding trajectory to achieve significant impact.
• “I think it’s increasingly clear to me that that this technology is is on a sort of compounding trajectory to do really massive things.” – Alpin Yukseloglu

Convergence of AI and crypto

• The convergence of AI and crypto is likely to be positive.
• “I think that for all the reasons we’ve talked about that for fundamental reasons crypto is extremely good for ai.” – Alpin Yukseloglu
• AI can be extremely beneficial for crypto if directed properly.
• “If we push things in the direction that we want them to go in that we can make ai be extremely good for crypto.” – Alpin Yukseloglu