Anthropic: 90% of Code Generated by Claude, Talent Retention in Question

Anthropic publishes ambitious internal indicators: in September, over 90% of code contributions were generated by Claude, and daily production per engineer in the third quarter increased by more than 45 times. However, the company remains heavily dependent on its leaders and key technical profiles. It emphasizes a "low-ego" culture designed to limit friction, while its data reports about 1% regrettable turnover over twelve months.
Dependence on Critical Profiles and Regrettable Turnover at 1%
Anthropic indicates that it relies heavily on its senior leaders and key technical members, warning that their departure could disrupt operations and slow down innovation. The company's internal data shows about 1% regrettable turnover over a twelve-month period ending on September 30. Anthropic specifies that if the convictions of these profiles change or if another organization seems better aligned with their goals, retention could become more challenging. In a growing industry where rare skills are highly sought after, this dependence poses an additional challenge. The company claims that many of these specialists are motivated by its mission and the resolution of specific AI problems. The sector is also experiencing departures accompanied by public alerts about the risks of increasingly powerful AI; researcher Jacob Coxon left Anthropic last month, believing that Anthropic and OpenAI were progressing too quickly towards a more powerful AI without a sufficient plan to control it.
Claude, a Lever for Productivity and Growth Without Linear Staffing
Anthropic states that it uses Claude to grow without a proportional increase in staffing, aiming for high individual productivity. The company notes that this tool enables the completion of tasks that previously would have required significantly larger teams. Most of its employees hold positions in research, engineering, or product development, and the number of collaborators has increased from 940 to 5,300 in less than two years, representing a more than fivefold increase. Anthropic believes that, without Claude, the workforce could have grown even more. In software engineering, employees increasingly oversee and validate the work generated by Claude instead of writing code line by line themselves. In September, Claude generated over 90% of the code produced by the company, and in the third quarter, the average daily contributions per engineer exceeded more than 45 times the daily average observed in 2024. Many tech players are looking to adopt this kind of method to enhance their agility.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
A Low-Ego Culture: Balancing Internal Promise and External Vigilance
Anthropic promotes a so-called low-ego culture, where the company's interest takes precedence over status, credit, or seniority. According to the company, this approach allows for an unusual distance between a research idea and a commercial product, reduces organizational friction, and increases impact per employee. Jamie Shapiro, an organizational psychologist and CEO coach, believes this approach is justified because relevant ideas can come from all members of the organization, including those who interact directly with customers. However, she notes that as Anthropic scales, it will become more complex to ensure that recruitment, recognition, advancement, and accountability processes continue to support these principles. In her view, the organization's growth complicates the task of ensuring that everyone adheres to the desired culture, which is determined by the least acceptable behaviors that are tolerated.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.