After a year of tracking crypto x AI, here's a braindump of things I've been seeing/thinking (tldr is we're waiting for a spark to send our tokies parabolic): 1. Most crypto x AI tokens are down 60-80% from their January peaks, with only a handful of newer projects showing sustained gains. Projects with defined use cases (i.e. $BNKR) are thriving (screw the X ban) while pure narrative plays (i.e. $GOAT) aren't. 2. AI Social Agents: The social AI agent boom has largely failed. Most "agents" are glorified chatbots flooding social media with low-quality content. The few exceptions (like Grok's contextual Q&A) succeed because they're transparently AI tools rather than human impersonators. The lesson: AI should augment human capabilities, not replace human authenticity. Either that, or LLMs get better at surfacing novel insights from the X firehose. 3. Agentic finance: Agentic portfolio managers on the other hand show genuine promise because they’re solving measurable problems: 24/7 monitoring, emotion-free execution, and strategy backtesting at scale. Projects like @Almanak__ (one we’re working w closely at Labs) are gaining traction not through hype but via yield. Tangential thought experiment: if AI eventually outperforms human traders, what happens to individual trading as a skill or profession? Edge gets eroded and it becomes -ev to be an unaugmented trader. There are also some interesting plays in the AI data space, but revenue is volatile and often obfuscated. We really need a decentralized data marketplace to democratize the playing field. 4. Development Tools: "Cursor for crypto" tools like @devfunpump represent genuine innovation. They’re lowering barriers to dApp creation and that could democratize crypto development. We're still waiting for the breakout vibe-coded crypto app that proves this thesis (I suspect it’s coming soon). 5. Funding Disconnect: The most compelling AI x crypto companies are raising private capital rather than issuing tokens. This creates a liquidity problem where retail investors chase low-quality tokenized projects while institutional money flows to higher-quality private deals. Until high-quality projects go public via tokens, broad repricing will be difficult. 6. Missing Catalysts: Someday, OpenAI will integrate Worldchain. Wish they’d just go for it now. I imagine there’s a lot of regulatory murk in the way, though. More realistic catalysts include: federal AI safety regulations that favor decentralized solutions (unlikely imo), major DeFi protocols or CEXes successfully integrating AI trading, or breakthrough improvements in on-chain inference capabilities. 7. Upcoming Unlocks: Decentralized AI models are rapidly improving and could unlock lots of use-cases we can’t yet see. Some Bittensor subnets are starting to generate actual revenue. x402 (shoutout to the @questflow team, which helped develop it) seems to be gaining steam as a leading agentic payments protocol. Near Intents showing promise. And I’ve been talking with the @ritualnet team, which is nearing public testnet. It uses “heterogeneous compute” (ie a mix of any form of compute rather than the simplified VMs most chains use) to let devs bring AI and offchain data onchain so smart contracts can utilize it/trigger actions. We need more flexible, futureproof platforms like this as innovation goes parabolic. 7. When to flip bullish: (1) OpenAI adds a Worldcoin integration or launches a social network with AI + payments integration, (2) clear evidence that on-chain AI inference can compete with centralized alternatives on cost/speed, (3) massive spike in assets under autonomous management and/or AI+CEX integrations, (4) some new unlock in capabilities. All the ingredients are here. We’re just missing the inevitable spark. When it arrives, it’s going to be epic.