Base abandons its social direction; the founder admits the strategic failure

Original｜Odaily Planet Daily（@OdailyChina）

_Author｜Wenser（__@wenser 2010 _）

Base, long regarded as the “king of L2,” has officially come to the end of its 3-year “Exploration Era.”

In the early hours of today, Base cofounder Jesse Pollak posted a message recapping the ecosystem’s development direction over the past two years, admitting that its earlier bet on on-chain-native social was wrong. He said that social tracks such as Farcaster, Zora, Miniapps, and creator coins failed to become a core driver of crypto adoption. As a result, Base also fell behind some competitors in areas such as perpetual futures, prediction markets, tokenization, and payments. In addition, he emphasized that responsibility for the Base APP would be handled by Cobie, a community leader whom Coinbase spent a large sum to complete a talent acquisition.

Earlier, Coinbase CEO Brian Amstrong also acknowledged, in response to community doubts, that “betting on content tokens was a mistake.”

Now in 2026, Base—having repeatedly flip-flopped on token issuance—has no choice but to face an awkward development period, which may once again validate a truth: No transactions, no crypto.

Base’s transformation: the founders go back to writing code, and the one-stop APP is handled by Coinbase’s team

Let’s first focus our attention on Jesse’s “mea culpa” post about the Base ecosystem.

In this long piece, Jesse first admits his mistake—although the bet on blockchain technology driving the success of crypto applications was correct, the choices made in specific sub-areas were wrong, as it mistakenly bet that social experiences would drive crypto applications.

Looking back, the Base ecosystem seems to have always been stuck in a kind of “self-entertainment” mode—

• On the leadership side, Jesse always mentions the “creator economy,” interacting with community members under the banner of the creator economy and encouraging all creators to issue coins and release NFTs, with the nice-sounding claim that “on-chain creator economics will rewrite the industry’s development.” But in reality, the on-chain creator ecosystem of Base is extremely hostile, and most creators’ earnings are even lower than those of developers who just jump on trending hype to issue coins on pump.fun;
• On the development strategy side, Base has long been perched under the banner of “Onchain Summer,” but in reality it has always followed behind Solana and BNB Chain with a follower strategy. When the Solana Meme coin boom came, Base played Meme coins too; when Chinese Memes on BNB Chain gained momentum, Base seized the opportunity to copy. Its collaborations with real-world brands such as McDonald’s have often remained superficial marketing initiatives, with no real action taken.
• In terms of trading activity, earlier data suggested that more than 80% of Base’s on-chain activity came from self-cycling by a small number of addresses, meaning the actual user base was extremely limited. Even in areas mainly promoted by Coinbase—such as x402 and AI Agent payments—transaction volume still has not been able to see a breakthrough on a truly standout scale, remaining in a closed loop of carving out its own niche.

As for the failures of SocialFi platforms such as Farcaster and Zora, it has long been evident to all. The former’s founder went on to work on payments at Tempo; the latter, after issuing its token, can be said to have “completed the task,” but its token’s drop has already exceeded 80%.

The beautiful vision Jesse imagined—“using social media to reach billions of people worldwide with crypto”—did not happen. After all, Web2 social media products are as abundant as there are cows, and not to mention, there are already a series of giant products that offer better user experience, more first-mover advantages, and richer creator content.

Many times, you think you’re playing Meme, coming up with jokes and memes, and activating a community—when in fact you’re building your own information cocoon, deliberately creating a small circle, and intentionally raising the entry barriers.

And the move of handing the ecosystem’s flagship product Base APP to the Coinbase team is more like pulling the Base ecosystem back into Coinbase’s embrace—accepting Coinbase’s unified arrangements and existing as part of the Coinbase ecosystem and as a component in a particular link. To a certain extent, this move expands the openness of Base APP, and naturally also means Base APP is no longer exclusive to the Base ecosystem.

For crypto projects, focusing on finance, trading, and investment remains the industry’s main storyline. Over the past two years, despite market fluctuations, the on-chain Perp platforms and prediction market platforms that became very popular are precisely hitting this trend. They can even benefit from highly volatile, high-risk market conditions to earn more protocol revenue and platform revenue.

In light of this, Base has finally figured it out. It is no longer fixated on “decentralized social” or “the creator economy,” but instead has shifted to three major directions: trading, payments, and AI Agents.

Base’s new chapter: focusing on trading, payments, and AI Agent development

Knowing one’s mistakes and making improvements—nothing is greater than that.

Although Base has taken many detours in the past, it now appears that Jesse, Coinbase CEO and others have seen the problems and also provided their own solutions.

Previously, Brian Amstrong responded to community doubts publicly by saying, “Base’s main business focuses on trading, payments, and AI Agent services (in that order). I think these three are closely connected. For example, to conduct payment transactions, you need foreign exchange resources; and agent services involve a large amount of trading and payment business. By the way, right now most resources are used for trading business. Maybe these businesses haven’t yet achieved external applications, but the actual situation is indeed like this.”

In today’s long post, Jesse also said that in the future Base will position itself as “a blockchain for global finance,” with 2026 focusing primarily on three directions: trading, payments, and AI Agents. Among them,

• Trading covers tokenized stocks, Meme coins, and application tokens;
• Payments revolve around global stablecoins for individuals and enterprises;
• AI Agents will use crypto as native computing money, serving large-scale participants in the future machine economy.

Base’s future business priorities are not just talk—they are the relatively best choices based on the current industry trends and its own strengths.

On the trading side, although Base’s active user base is limited, its advantages—such as lower Gas friction costs, the Ethereum ecosystem’s network effects, and the compliance-oriented Coinbase U.S. background—can ensure it maintains advantages and value at the foundational construction level in the development of tokenized stocks, Meme coins, and serious projects;

On the payments side, Base’s low interaction costs, the underlying x402 protocol foundation, and Circle’s USDC support also lay the groundwork for both individual payments and institutional payments, as well as stablecoin adoption. Combined with a one-stop application like Base APP, if it can later connect more supply-side brands, merchants, and institutional users, it is expected to promote large-scale stablecoin adoption in the U.S. and even around the world.

On the AI Agent services side, this step is even more like an early move toward the future AI Agent economy. Although AI Agents are still in an early stage of development, judging from news such as OpenAI’s GPT 5.6, Anthropic’s Fable 5 model, and recent reports that Apple’s AI scheduled-collaboration partner includes Alibaba’s Qianwen model, it suggests that the day when AI Agents use crypto for widespread payments, trading, and consumption will come soon. Earlier, according to Alibaba’s official announcement, as of this May, the number of Alipay’s AI autonomous ordering and payment transactions had already reached 300 million (covering scenarios such as AI agent payments, AI ordering, AI consumption, etc., including payments completed through AI applications such as Qianwen and Rokid).

Of course, the shift in Base’s strategic direction is also accompanied by personnel changes behind Coinbase.

Coinbase organizational reshuffle: some people leave the circle, while others keep writing code

Previously, Coinbase engineering lead Brock Miller said publicly that he had left and joined Anthropic to become a member of its technical team. When crypto is no longer a hot trend, some people choose to go along with the tide and join the AI revolution wave.

In early July, Coinbase Chief Legal Officer Paul Grewal announced his departure and said he would go work at a startup company afterward. He added: “I will continue to serve as an adviser to Coinbase, and participate in work related to Coinbase trust charter through the U.S. Currency Regulator.” After that, Coinbase announced that Molly Abraham would lead the company’s legal team as General Counsel, Ryan Van Grack would become Vice Chairman, and is expected to take on broader and more public-facing responsibilities.

This is an example of someone leaving but still maintaining a cooperative relationship with Coinbase.

As for the people who write code—it is none other than Base’s founder Jesse—who said himself: “I’ve started rewriting code again, and I’ve launched some products.”

Of course, these departures or changes may not be entirely a reflection of personal willingness; they may also be driven by the impact of AI Agent development.

Recently, Coinbase platform head Rob Witoff stated that currently more than 95% of the company’s code is written by AI or completed with the assistance of AI, representing a significant increase from the 40% proportion announced earlier this February. In an interview, he said: “In fact, 100% of Coinbase employees use AI every day.” He** also said that “most Coinbase engineers are running 5 to 10 AI Agents at the same time. The combined work capabilities of these AI Agents are equivalent to about 1,200 employees.”

He expects that “by 2030, Coinbase’s AI Agents might take on a workload equivalent to 100,000 employees. However, he said that key areas such as core cryptography still require human involvement, and AI is mainly used for code testing, vulnerability checks, and prototype development.”

It’s not only the Base ecosystem that will serve the AI Agent economy in the future—perhaps in the future most of Coinbase’s employees will also be replaced by AI Agents.