Is Claude’s Momentum Sustainable?

Boris Cherny often finds himself in the spotlight, whether at a bar, an airport, or any public venue, where fans eagerly request selfies with the mastermind behind Claude Code.

In recent months, Anthropic’s Claude and its coding platform have surged in popularity, capturing attention across social media, engineering communities, and executive offices. Over the holiday season, Claude Code gained traction as users created everything from tools for analyzing MRI results to alternatives for Goodreads, and even an AI-driven T-shirt design contest with a unique judicial framework. By January, social media posts were buzzing with claims that Anthropic was “taking the lead” over OpenAI, speculating on the model’s capabilities and whether this signaled a decline for OpenAI.

Between late December and late January, Anthropic’s visibility skyrocketed by 13 percentage points, while OpenAI’s presence saw a slight dip, according to data from a stakeholder intelligence platform. By November 2025, Claude Code had surpassed $1 billion in revenue, and the company was reportedly negotiating to double its latest funding round to $20 billion, reflecting a surge in investor interest.

Now, Anthropic is set to unveil a new model, Opus 4.6, which the company touts as a “direct upgrade” from its predecessor, boasting enhanced speed and accuracy for a variety of tasks, from coding to crafting presentations in Excel and PowerPoint. Dianne Na Penn, Anthropic’s head of research product management, noted that Opus 4.6 represents an “evolution from past models,” allowing it to “think” longer and tackle more complex questions.

After riding the wave of hype surrounding Claude Code, Anthropic aims to solidify its position with the launch of Opus 4.6. But can the company sustain its momentum post-holidays?

“I think we’ll remember December 2025 as a turning point where everything changed,” said Mike Brevoort, principal architect at a warehouse automation startup. He recounted testing Claude Code and other tools during the holiday break, recalling a moment of realization when he used Anthropic’s model to build a prototype: “Wait a minute, that’s possible?” He believes “Claude is becoming the verb now, or the noun, in the same way ChatGPT did at its launch. It’s been an interesting shift.”

While tech professionals across various sectors were busy utilizing Claude Code, Cherny was on a houseboat in Copenhagen, also leveraging Claude Code to enhance its own capabilities. With over five Claude agents operating in the cloud on his behalf, Cherny submitted more than 300 pull requests in December, marking his most productive month at Anthropic in his year-and-a-half tenure (excluding a brief stint at another company). Since November, he has relied on Claude Code for all his coding needs, a tenfold increase from the previous year. Across Anthropic, Claude Code reportedly handles 70 to 90 percent of coding tasks, with about 90 percent of its own code generated using the tool.

Claude Code first launched in February 2025, and its recent rise to fame has been a long time coming, according to numerous tech industry leaders and engineers. Anthropic revealed that in the past year, it has seen an eightfold increase in business customers generating over $1 million in revenue. The company also outperformed competitors in multiple categories at a recent awards event, including “best agentic model” for its performance on popular leaderboards.

However, the rapid adoption of Claude Code over the holidays caught many by surprise. “The shift from various coding interfaces to Claude Code has been astonishing,” remarked Maggie Basta, a partner at a venture capital firm. Part of this success can be attributed to a holiday promotion that doubled rate limits for certain users, attracting new and casual users during the break. “For those who don’t engage with this daily, the transition to ‘Oh wow, this is impressive’ was significant,” said Austin Parker, director of open source at a tech company. Much of Claude Code’s success can also be traced back to the release of the Opus 4.5 model just before Thanksgiving.

Opus 4.5 was seen as a game changer, enabling Claude to tackle more complex tasks over longer periods with less guidance, resulting in significantly improved outcomes.

Others echoed this sentiment. “It shifted from needing constant oversight to succeeding by default,” said Josh Albrecht, CTO of an AI company, who made such progress that he didn’t even try OpenAI’s competing Codex during the holidays. Albrecht noted that his team has revamped its processes to give small groups of engineers nearly unrestricted access to Claude Code credits, allowing them to see how quickly they can deliver results.

“I find that I have to describe my intent a lot less,” said Allie K. Miller, CEO of an enterprise AI advisory firm. “Before Opus 4.5, you had to break things down step by step. Now, it’s more like, ‘Here’s the magic castle. Build it.’ And it gets done.” This enthusiasm is echoed across various industries.

Ran Liu, chief AI scientist at an education startup, mentioned that her entire team prefers Claude Code over other platforms. At a healthcare-focused AI career platform, the majority of AI-generated code is produced using Claude, according to the CEO. Andrew B. Hall, a professor at Stanford, is currently experimenting with Claude Code to see how effectively it can update old research papers, with human oversight for accuracy.

For Siqi Chen, CEO of a finance analytics firm, the product’s most significant advancements involve its ability to self-organize and handle long-term tasks, even complex, multistep projects over several days. He noted the shift from simple conversations to having agents perform highly productive work.

Overall, sources agree that Claude Code’s success reflects broader advancements in AI agents, which are designed to complete intricate, multistep tasks autonomously. Each year, AI companies promise that their agents are ready for prime time, and while past efforts have fallen short, recent developments, particularly in coding agents, show significant progress.

For some, Anthropic’s coding agents represent a glimpse into the future of autonomous agents. “Claude has emerged as a clear example of what an autonomous long-running agent can achieve, starting with coding but clearly extending into other areas of knowledge work,” said a tech CEO. He believes Anthropic is well-positioned due to Opus 4.5’s success in tool utilization, reasoning, and complex data analysis, emphasizing that allowing a tool to handle tasks that would take humans hours or days requires a level of sophistication that has only recently been achieved.

While Opus 4.5 has dominated leaderboards, real-world adoption is the ultimate goal, especially since companies can tailor models to excel in specific benchmarks. Even during the peak excitement surrounding a competitor’s release, many in the tech industry expressed their intention to continue using Anthropic models for coding, primarily due to their preferred user experience.

Hype tends to be cyclical, and Anthropic’s rivals are keen to reclaim their competitive edge. Recently, a competitor’s CEO hinted that other tools might require “more handholding” and that their latest model is superior for complex tasks. This week, a new standalone app was launched to capture more of the AI coding market.

The question remains whether Anthropic can maintain its position with the launch of Opus 4.6.

One potential advantage for Claude is its ability to foster user loyalty. AI companies are currently focused on creating “moats” through personalization and memory features, alongside subscription models. Claude Code was among the first AI coding services to offer a customizable file with basic operating systems and instructions, which users can modify over time. One user reported having 1,000 such instructions, making it cumbersome to switch to another platform.

Steve Croce, field CTO at a Python platform for AI and data science, believes that existing users are unlikely to switch unless a significant reason arises. “Users stick with tools they trust and have seen success with,” he said, noting that his engineering team predominantly uses Claude Code due to its favorable user experience. “At this point, they’re not going to change unless something drastic happens, either from OpenAI or Google, or if Anthropic gives them a compelling reason to leave.”

Brand recognition also plays a crucial role. One CTO compared Claude and Claude Code to Apple’s success with the iPhone and iOS. Some users who prefer a competitor’s model still acknowledge Anthropic’s early market entry and its tendency to roll out features ahead of others. Runway’s Chen noted that Claude Code was the first to grant an AI agent “raw control over the fundamental tools of your computer,” while many competing tools operate solely in the cloud.

As Anthropic launches Opus 4.6, it will encounter numerous challenges in maintaining its lead. Even before the update, several users expressed interest in open-source alternatives to Claude Code due to cost considerations. Additionally, public trust in the company has waned, with recent data showing a decline in Anthropic’s trust and likability scores, alongside a more significant drop for a competitor.

Concerns about security also lingered with Opus 4.5. An AI researcher noted that while it excelled in clarity and conciseness, another model was deemed the most secure, exhibiting fewer vulnerabilities. “Specific types of bugs have worsened in Opus,” he stated, attributing this to the model’s complexity. In response, Anthropic plans to address these concerns in Opus 4.6’s final model card, having invested in enhanced cybersecurity measures.

Anthropic has also cultivated a reputation as a responsible player in the AI space, distancing itself from controversial associations. Users have expressed dissatisfaction with a competitor’s recent controversies, leading some to seek alternatives like Claude and another model.

“ChatGPT lost its appeal a while ago,” said Croce. “There’s been a lot of distracting noise from that company … Anthropic doesn’t create unnecessary noise for itself.”

Like other AI firms, Anthropic faces pressure to innovate rapidly while maintaining ethical standards. However, it has focused on marketing text-based AI productivity tools to enterprise users, avoiding the controversies that have plagued other companies. Anthropic’s Penn stated that Opus 4.6 continues the company’s commitment to making Claude a valuable tool for work.

This could be reassuring for many users. “I don’t have to worry about Claude making controversial statements or engaging in bizarre behavior,” said Parker, referring to issues faced by other AI models. “We want to avoid brand risk, and there’s a lot of that in the generative AI landscape.”