Grok Build Under Fire: Developers Compare xAI's Tool to Claude Code and Codex
Elon Musk recently posted a request on social media platform X for criticism of his new product, Grok Build, in response to an enthusiastic post from one user. The community's reaction was far from optimistic. Dozens of comments appeared under the billionaire's post, and a significant portion pointed out the tool's weaknesses. Let's figure out what's wrong with Grok Build and why, according to Musk's followers, competitors like Anthropic and OpenAI have pulled far ahead.
What is Grok Build and what is its strength?
Grok Build is an agentic CLI tool from xAI for development, launched in early beta in May 2026. It works directly from the terminal and is only available to SuperGrok and X Premium Plus subscribers. The access cost is around $300 per month, placing the product in the same price range as Claude Code and GitHub Copilot.
For complex tasks, the developers have included a planning mode: the user can approve the plan, comment on individual steps, or completely rewrite it before execution begins. After that, each change is displayed as a diff (a line-by-line display of the differences between two file versions). The base model, Grok 4.3 beta, uses a 16-agent architecture and a context window of 2 million tokens, and the tool can run up to eight parallel agents. It sounds impressive, but in practice, things turned out differently.
Comparison with competitors is not in xAI's favor
The most frequent topic in the comments was comparison with Claude Code and OpenAI Codex. Unfortunately, these comparisons almost always turned out to be unfavorable for Grok. One developer described a direct experiment: Grok worked on implementing a project for almost two days, after which the same volume of tasks was switched to Codex. As a result, the competitor advanced twice as far in six hours.
Another user stated that Grok went into infinite loops for thirty minutes, while Opus solved the same problem on the first try. A third specialist noted that the inference speed in Grok CLI feels too slow compared to Claude Code and Codex, making it uncomfortable to watch the agent work.
Additionally, there were opinions that Grok is good for deep research but clearly lags behind competitors in complex autonomous coding. This is a serious signal: if the tool cannot handle basic development tasks, its value to the professional community drops sharply.
Feature requests and desktop application
A notable group of feedback concerned the system's missing capabilities. Users actively requested the creation of an official desktop application, similar to Claude Cowork. They noted that Claude's main strength lies in its integration into all aspects of workflows, not just in writing code.
Additionally, requests were made for the following options:
- releasing an open-source version of the product during the beta testing period;
- implementing full loop skill functionality;
- creating a /goal command for stable autonomous agent operation;
- integrating a built-in capability to demonstrate the software being created without exporting.
The issue of feedback was also raised separately. One user admitted they didn't understand through which channels to send feedback after an unsuccessful result. This is quite notable, given that xAI embedded the /feedback command directly into the CLI. It seems that even basic user interaction mechanisms remain unclear.
Price and limitations: economics vs. developers
The subscription cost sparked a separate wave of criticism among specialists. Users complained about the strict tie to the expensive SuperGrok plan and suggested introducing a more affordable tier. Additionally, they lamented the strict token limits and the 15-minute daily access limit to Grok Premium.
In the table below, I have summarized the main complaints:
| Problem | Essence of user complaint |
|---|---|
| High entry barrier | Mandatory tie to the expensive SuperGrok plan |
| Usage limits | Token restrictions and only 15 minutes of Grok Premium per day |
| Geography | Fears that xAI will repeat Claude's withdrawal from Europe |
Irony and skepticism: how the community reacted to Musk's post
The very format of Musk's post drew sarcastic reactions. Several users pointed out that the request for critical feedback was accompanied by quoting an enthusiastic fan who literally professed love for the product. One commenter called reposting one's own praise a special kind of self-confidence.
At the same time, part of the audience remained loyal to the company. There were thanks for the team's fast iterations and statements that the product is rapidly improving. Some even predicted that Grok would soon become the best tool on the market. However, such voices were clearly in the minority.
Cryptalist analyst conclusions
The collection of feedback under Musk's post revealed an obvious gap between the marketing message and the assessments of practicing developers. Grok Build stands out with its large context and multi-agent architecture, but in real-world tasks, users note it lags behind established Claude Code and Codex in terms of autonomous coding quality, speed, and stability.
Thus, the key complaints boil down to three areas:
- Quality and reliability: loops, regressions, and losing to competitors in direct tests.
- Ecosystem: lack of a desktop application, open-source version, and several agentic features.
- Economics: high price and strict token limits.
The early beta stage and built-in feedback mechanism give xAI a direct channel for quickly refining the product. However, in my opinion, to catch up with the leaders, the company will need not just to fix bugs, but to radically rethink its approach to autonomous coding and user experience. For now, Grok Build is a promising but raw tool that loses to more mature solutions on all key metrics.