Tinygrad abandons bounties as AI gaming breaks hiring signal

Overview

Tinygrad, the open-source artificial intelligence framework founded by George Hotz, has officially retired its bounty program, citing the pervasive impact of generative AI on hiring practices. The decision marks a significant shift in how the company identifies and recruits talent, moving away from transactional bug-bounty models toward a more organic, contribution-based evaluation process.

Hotz explained that the bounty system, which previously served as a reliable signal for engineering skill and problem-solving ability, has been rendered ineffective by automated tools. In an era where candidates can paste incomplete task descriptions into AI coding assistants like Claude Code and submit pull requests without truly understanding the underlying request, the filter no longer distinguishes genuine expertise from AI-generated output.

Consequently, Tinygrad is pivoting to a model where prospective hires simply begin working on the project. By engaging with the codebase, participating in public Discord channels, and attending open weekly meetings, applicants can demonstrate their value through sustained, meaningful contribution rather than one-off submissions.

The Death of the Bounty Signal

The core issue driving this policy change is what Hotz described as an "arms race" in the hiring pipeline. Traditional automated resume screening systems are now being countered by automated resume spammers, creating a stalemate that wastes the time of both employers and job seekers. The bounty program was originally designed to bypass this noise by offering tangible rewards for specific technical tasks, but the advent of capable AI coding agents has broken this mechanism.

"Now, it's people who barely even understand what the bounty is asking for pasting it into Claude Code and submitting a PR. The bounties used to provide signal, now they don't. So that chapter is kind of over."

According to the company, the barrier to entry for completing bounties has dropped so low that the resulting pull requests rarely reflect the deep understanding or code quality required for the role. This saturation means that completing a bounty no longer guarantees that a candidate possesses the critical thinking or architectural skills Tinygrad seeks.

New Hiring Methodology: Contribute First

In lieu of bounties, Tinygrad has adopted a "show, don't tell" approach to recruitment. The company emphasizes transparency, with all coordination occurring through public Discord servers and GitHub repositories. Weekly meetings are open for anyone to listen to, and archives are available for review, allowing candidates to fully immerse themselves in the company’s technical culture before applying.

Candidates are encouraged to take their time to understand the project's goals and identify areas where they can help. Hotz noted a distinct preference for contributions that prioritize code deletion and simplification over the addition of new features. The team values deep bug fixes and comprehensive regression tests over superficial additions.

To succeed under this new model, applicants should:

1. Read the Discord history to understand current debates and technical directions.
2. Study the commit history to grasp the project’s evolution and coding standards.
3. Engage with the community by attending open meetings and contributing to discussions.

This method ensures that new hires are deeply aligned with the project’s ethos and have already proven their ability to collaborate effectively within the existing infrastructure.

Philosophy on AI and Software Development

While the bounty program suffered due to AI, Hotz pushed back against the notion that these tools fundamentally change the definition of a software engineer. He acknowledged that AI can generate "slop apps" for "slop brands" cheaply, comparing such work to using a slightly more flexible version of WordPress. However, he argued that this does not constitute true software development, particularly at the infrastructure level where Tinygrad operates.

"We encourage use of AI in software development workflows in the same way we encourage other tooling, but there's nobody now that we would hire because AI exists who we wouldn't have hired before and vice versa."

The company maintains that AI is merely a tool, and the fundamental requirements for hiring remain unchanged. The challenge lies not in the nature of the work, but in finding new, robust ways to screen for talent in a landscape where low-effort automation is abundant.

Strategic Vision and Logistics

Beyond hiring, the announcement touched on the broader strategic mission of Tinygrad. Hotz framed the project as a long-term play, spanning a ten-year horizon, to attack what he views as big tech rent-seeking behavior. By commoditizing petaflops—the raw computational power sold by major tech companies—Tinygrad aims to remove the infrastructure layer that allows larger firms to extract excessive value.

"If the tech companies sell repackaged petaflops and we commoditize the petaflop, there's no place for them to rent seek."

Logistically, the role is described as highly flexible. It is a remote or hybrid position with offices in San Diego and Hong Kong, though employees are free to work from either location or entirely remotely. The commitments are minimal, requiring only a one-hour weekly meeting and a brief three-sentence sprint summary every other week.

Compensation ranges from $75,000 to $150,000 annually, plus equity ranging from 0.1% to 0.5%, depending on experience and scope. With almost all work being open source and public, the barrier to entry is not about navigating corporate opacity, but about demonstrating genuine technical merit through visible, sustained effort.