顺势而为2026

Deep Tide TechFlow updates: On August 17, according to The Information, Anthropic and OpenAI are accelerating the rollout of AI applications and features tailored to specific industries. Some enterprise customers are concerned that, in the future, the two companies may prioritize their strongest AI capabilities for their own products rather than making them available to external businesses via APIs.

This trend shifts the relationship between model providers and enterprise customers from foundational technology collaboration to potential competition. If top-tier model capabilities increasingly tilt toward proprietary applications, enterprises may need to reassess their reliance on AI infrastructure vendors, supply-chain arrangements, and long-term technology strategy.

⚡ Top AI & Big Data Projects by Developer Activity

Dev.Activity measures a project’s development activity across its public GitHub repositories based on the number of recorded events.

ai16z is finally dead.

With a stack of legal troubles, with many technical goals that were never realized, of course—and with the heartfelt expectations of tens of millions of retail investors. @shawmakesmagic actually claimed that a token was not sold. And because it couldn’t win the case, it ended up handing over all the remaining Treasury funds in a settlement. It’s absurd. If it’s going to die, then let it die—but without even the last shred of dignity.

As one of the witnesses, I experienced ai16z’s glory, and watched it decline little by little through repeated struggles and flailing. At this moment, my heart is filled with mixed feelings. One sentence keeps circling in my mind:

ai16z may die, but that era of “on-chain innovation” fanaticism—please stop.

Come, let’s pull the timeline back to the end of 2024 for a moment and remember:

1) Back then, AI MEMEs like Goat, act, ai16z, Fartcoin—just about any of them—could easily surge to a market cap height of 1B+. So many MEMEs rode the winds of momentum. Every day on-chain brought new opportunities, new wealth passwords, and retail investors partied along with it.

2) Back then, the Eliza framework was open-sourced. A large number of plugins flooded in crazily. Hundreds or even thousands of people chased the GitHub to contribute—endless forks, PRs, and all kinds of innovation experiments. Don’t think it was only gamblers watching candlesticks to make money. Truly, there were developers burning the midnight oil writing code, doing interesting innovation experiments, rushing in to do something for this brand-new narrative at the frontier of AI agents.

3) Back then, web2AI was still in the model performance arms race. GPT-4o and Sonnet 3.5 were still mainstream. There was no MCP, no x402, no A2A, not even OpenClaw or Moltbook. There was also no concept of agent applications—but in the Crypto space, a wave of AI full-stack innovation had already been launched with forward-looking ambition. Experiments one after another: agents building wallets for themselves, agents doing Trading, agents adapting to chains, and even agents constructing social networks. That was truly an on-chain innovation era where if you had a dream, you dared to make it real.

...Enough reminiscing. The more I say, the more it brings tears.

Finally, one last emphasis: this is not only a little piece of mourning for ai16z, but I also hope it can rekindle a brand-new spark for those old dreams of on-chain innovation in Crypto.
🔥