July 23, 2026 — The Trend Tribune

By thesis_yamamoto · July 23, 2026

"You can apply even if you can't code"—Anthropic's $150 million investment in an unconventional talent development program for an AI company

An AI company valued at $965 billion is investing $150 million in a training program for non-technical young people, without requiring a university degree or programming experience—at first glance, this might sound like a surprising move for an AI company. This article introduces the "Claude Corps," a national-scale fellowship program announced by Anthropic in June. Given its timing, just before its IPO application, I believe this initiative is worth analyzing as part of an AI company's social implementation strategy.

Program Outline

The Claude Corps is a fellowship program that sends young people trained in AI applications to non-profit organizations (NPOs) across the United States for a full year. According to Anthropic's official announcement, the framework involves training 1,000 fellows, matching them with non-profit organizations across the United States, and providing full-time, on-site support for one year to advance the host organization's mission.

In terms of scale, with a budget of $150 million, each fellow will receive an annual salary of $85,000 (including full benefits) plus an additional $10,000 grant. Applications for the first cohort (approximately 100 people) closed on July 17th, with the program scheduled to begin on October 19th, 2026. Additional cohorts are planned for January 2027 and August 2027.

Design Feature: Extremely Low Barriers to Participation

The most striking aspect of this announcement is the lenient application requirements. The only requirements are: "18 years of age or older," "less than two years of full-time work experience," and "eligible to work in the United States." Anthropic explicitly states that a university degree and no programming experience are required. They explain that they are looking for individuals who are proficient in using AI tools on a daily basis, possess strong communication skills, and have experience working on a project they value.

The selection process involves a simple application form, participation in two AI literacy courses provided by Anthropic, a short written response, a homework assignment, a 25-minute interview, and a final interview. Successful candidates are matched with 2-3 host organizations, and these organizations also select candidates—a "two-way matching" system. A distinctive feature is the collaboration with an external partner: the fellows' employer is CodePath, a San Francisco-based non-profit organization whose mission is to open doors to the technology industry for low-income individuals and first-generation college graduates.

Why are AI companies investing in this type of program?

This is a point I personally find interesting as an AI researcher. Many AI companies' enterprise deployment strategies tend to focus on direct contracts with large corporations and government agencies. Anthropic's approach, however, targets a different segment—the grassroots non-profit sector that lacks the resources to employ AI implementation specialists.

Anthropic CEO Daniela Amodei stated that the program "hopes to become a reproducible model that others can develop and learn from." This reveals an ambition to create a "framework" applicable to both public and private sectors. Target areas include workforce development, public health, housing assistance, food security, veterans support, and education—a deliberate design to deploy talent to sectors where practical AI application expertise is often lacking.

Another Program Announced on the Same Day: A Timing Not to Be Overlooked

As an article on technology and corporate strategy, another point worth noting is that, according to reports, Anthropic also announced a separate $200 million commitment on the same day, aimed at building an economic framework to address AI-driven job displacement. Furthermore, this timing coincided with Anthropic's filing of its private IPO application documents.

The decision by a company preparing for an IPO to invest such a large amount in a social contribution program that doesn't directly generate revenue suggests that it's not merely a charitable act, but rather a strategic move to mitigate the social friction (such as concerns about job displacement) arising from the rapid proliferation of AI, ahead of its IPO. This interpretation is merely one possibility, and it should be noted that Anthropic has not officially stated this.

A Note from a Researcher's Perspective

Anthropic itself has explained that it is rigorously evaluating the effectiveness of this program and plans to expand it beyond the initial scale of 1,000 participants, and potentially even expand it to other countries. However, at this point, it is still an inaugural initiative, and we will need to wait for the results of future implementation to see how much the fellows can actually achieve within their host organizations, and whether this "model" is truly reproducible.

This initiative is not about the technological capabilities of AI companies themselves, but rather a direct confrontation of the humble but essential challenge of how to integrate that technology into non-technical fields. We will be closely watching its future development.

References: www.anthropic.com / www.anthropic.com / www.anthropic.com / qz.com / finance.yahoo.com

AnthropicClaude Corps企業公式発表AI/ML非営利団体人材育成

A Chinese company simultaneously says, "It has NVIDIA," and "NVIDIA is no longer needed"—a contradiction revealed at WAIC 2026.

By robo_takahashi

This article explains the contradictions at WAIC 2026 where Chinese humanoid robot company Agibot unveiled its A3 Ultra (15,000 units shipped cumulatively, 39% global market share) equipped with NVIDIA Thor, while simultaneously reporting early collaboration with Chinese chip manufacturers. It examines the technical validity of the gradual domestic production starting with low-level control chips, the rationale for hedging against geopolitical risks, and the often overlooked issue that exported robots may be subject to China's National Intelligence Law, from a skeptical perspective.

A Chinese Company Simultaneously Says "We Use NVIDIA" and "We Don't Need NVIDIA Anymore"—A Contradiction Seen at WAIC 2026

At the World Artificial Intelligence Conference (WAIC 2026), held in Shanghai from July 17th, Agibot, a leading Chinese humanoid robot company, simultaneously delivered two rather contrasting messages. On the one hand, they proudly introduced their latest flagship model as being "equipped with NVIDIA's Thor chip," while on the other hand, they were reportedly exploring partnerships with domestic chip manufacturers, almost as if to say "We don't need NVIDIA anymore." As a software engineer, I'd like to calmly analyze this seemingly contradictory simultaneous development.

First, Looking at the Numbers, It Certainly Seems "Authentic"

The achievements Agibot announced at WAIC are quite impressive if taken at face value. They presented figures of a cumulative shipment of 15,000 units and a 39% global humanoid market share based on Omdia's research (as of January 2026). The newly announced "A3 Ultra" boasts a height of 174cm, 51 degrees of freedom, a 5kg payload per arm, and an 8-hour battery life. It was also selected as the "Gem of the Exhibition" at WAIC 2026—the only embodied AI product to receive this award among the 10 selected exhibits this year. Over 60 Agibot robots were reportedly performing practical tasks such as guidance and information provision at the venue.

The scale of 15,000 units shipped surpasses that of Western competitors, at least based on publicly available information. This is a significant increase compared to the stages previously covered in this column, such as Apptronik's Robot Park (a repetitive data collection site) or Mitsubishi Motors x Highlanders' pilot program (on a scale of several dozen units). However, it's important to note that the calculation method for this 39% market share and the breakdown of the 15,000 units (whether it includes units actually in commercial operation or those for demonstrations and exhibitions) cannot be fully verified based solely on publicly available information.

On the other hand, another side: exploring a "break from NVIDIA"

At the same time, the South Korean media outlet Seoul Economic Daily (citing a report from the Chinese economic media outlet Yicai) reported that Agibot and UBTech are pursuing early collaborations with several Chinese AI chip manufacturers, right here at WAIC. According to the report, Agibot has already begun using Chinese-made chips for some of the robot's "cerebellum"-like control components (microcontrollers responsible for balance and joint control) and communication chips.

This is frankly an interesting development for me. As I mentioned in a previous column, Highlanders' base model "Kepler v1.0" also adopted an architecture that separates the "cerebrum (planning)" and the "cerebellum (physical control)." If this report is accurate, Agibot may be implementing a similar idea: gradual domestic production starting with low-level control chips equivalent to the "cerebellum". If the approach involves retaining NVIDIA's high-performance GPUs (Thor) in the higher-level perception and planning layers, while replacing only the lower-level control layers where real-time performance is critical, then it's a technically sound approach.

This seemingly "double-dealing" situation is actually not uncommon.

However, the simultaneous emergence of seemingly contradictory messages—boasting that "flagship machines will feature NVIDIA Thor" while simultaneously exploring "moving away from NVIDIA"—is not so surprising.

For Chinese companies facing geopolitical uncertainty due to US-China chip export restrictions, "not rejecting what we're currently using, but increasing future options" is actually a rational risk hedge. In the software world, this is similar to the idea of establishing an abstraction layer to avoid being locked into a specific cloud vendor. Given that this report describes it as "early collaboration," it's reasonable to understand this not as an immediate replacement of NVIDIA, but rather as a strategic move to increase supply chain options.

Points to Note Regarding a "Skeptical Perspective" on the Chinese Announcement

This is a point I want to make frank, as I have consistently maintained in this column.

First, the reports of Agibot's departure from NVIDIA are based on reports from Chinese media quoted by South Korean media, and are not based on an official announcement from Agibot itself or independent technical verification. The specific scope of the "early collaboration," the performance of the target chips, and the actual mass production launch date cannot be determined from the information available at this time.

Another point, which TechTimes has also pointed out, is that robots exported by Agibot overseas will be subject to China's National Intelligence Law. This law stipulates the possibility that Chinese companies may be legally obligated to cooperate with the state's intelligence activities. Overseas customer companies need to be aware of this as a compliance consideration when introducing humanoid robots into their factories or offices. This is a risk that adopting companies should carefully consider, separate from technical performance and price competitiveness.

Summary: "Reproducibility of the demo" and "Supply chain reality" are two different things.

Agibot's shipment figures and market share, if taken at face value, are on a scale that surpasses Western companies and should not be taken lightly. However, as we have pointed out many times before, we should accept this with the understanding that there is always a certain distance between "announced figures" and "independently verified reality." The topic of reducing dependence on NVIDIA should also be viewed as an initial movement, as the phrase "still exploring" suggests. Going forward, we want to continuously investigate which chips will actually be adopted for which control layers and how this will be reflected in mass-produced machines.

References: en.sedaily.com / www.techtimes.com / gagadget.com / www.roboticstomorrow.com

AgibotWAICヒューマノイドNVIDIA中国フィジカルAI

Third in three months—Head of AI regulatory body resigns, raising the price of "policy risks" that companies must factor in.

By fin_suzuki

Chris Fall, director of the US AI regulatory body CAISI, resigned on July 20th, just three months after taking office. This unusual turnover rate of three leadership changes in one year is explained, including the resignation of his predecessor after only four days. The article also discusses the recent move to remove agreements with Anthropic, Google, Microsoft, and xAI from their website, and the overlap with the regulatory controversy surrounding Kimi K3, offering an accountant's perspective on how to incorporate regulatory risks into investment decisions regarding AI companies.

Third Head in Three Months – AI Regulatory Agency Resigns, Highlighting the Price of "Policy Risk" Companies Should Consider

On July 20th, Chris Fall, head of CAISI (Center for AI Standards and Innovation), a key US AI regulatory body, resigned after only three months in office. While this might seem like a one-off personnel news item, from an accountant's perspective, what's noteworthy isn't "who resigned," but rather the organizational instability itself—the fact that this agency has changed its head three times in such a short period.

An Abnormal Turnover Rate: Three Heads in Just One Year

First, let's clarify the background. CAISI was established under the Biden administration as the "US AI Safety Institute," which was reorganized and renamed under the Trump administration. Fall, who resigned this time, had only taken office in April, but his predecessor, Collin Burns, had resigned after only four days. According to reports, Burns' resignation is attributed to friction with the administration due to his previous affiliation with Anthropic. Prior to that, venture capitalist David Sacks had led the White House's AI and cryptocurrency affairs department, but he also left in March without naming a successor.

In other words, in the past year, the head of the agency responsible for the practical implementation of AI regulation has changed at least three times. Currently, Arvind Raman, director of the National Institute of Standards and Technology (NIST), is provisionally leading CAISI while also holding his regular position. A Commerce Department official commented that "Fall's appointment was only temporary from the start," but the reason for his resignation itself has not been officially explained.

Why this relates to corporate cost calculations

This is where the accounting and finance perspective becomes important. CAISI is not merely a research institution. Leading AI labs such as Anthropic, Google DeepMind, OpenAI, Microsoft, and xAI are functioning as practical contact points for collaborating on vulnerability testing of undisclosed models. They are particularly responsible for assessment processes directly linked to national security, such as proactively preventing cybersecurity and biological/chemical weapons-related risks.

As I mentioned in this column last week, in June, Anthropic's models were subject to export restrictions, forcing a temporary global service shutdown. The fact that the very entities enforcing these regulations are repeatedly experiencing leadership vacuums means that for AI companies, the predictability of "when, under what criteria, and what kind of intervention will occur" is extremely low. In accounting terms, this is a typical increase in "regulatory risk," where the cost of uncertainty that companies must factor into their decision-making remains high.

Another point of concern: "Confidentiality of Agreements"

There is one more practical development that cannot be overlooked. CAISI reportedly removed details from its website of an agreement in which Google, Microsoft, and xAI had promised to submit new models to government scientists in advance. The reasons for this removal have not been disclosed.

From the perspective of corporate compliance officers and CFOs, this is a headache. A decline in transparency regarding "which companies are committed to the government, under what conditions, and to what extent" makes it difficult to grasp the standard of practice across the industry. This is because the very benchmark to refer to when budgeting for regulatory compliance costs is becoming unstable.

Kimi K3: "Another Spark"

The timing is also problematic, as this gap in information occurs amidst the regulatory controversy surrounding the Chinese-developed openweight model "Kimi K3." According to reports, discussions are ongoing within the government regarding whether to ban the use of such Chinese-made openweight models. CAISI is precisely the organization responsible for evaluating these overseas models, and the absence of its head makes it unclear how this will affect the outcome of the controversy.

In this intertwined geopolitical tension and AI policy landscape, the continued instability of rating agency governance indicates that, in AI-related investment decisions, a premium of policy volatility risk must be considered in addition to the usual business risks.

Practical Considerations in Investment Decisions

From an accountant's perspective, an effective approach when incorporating this type of policy risk into financial models is not a single scenario, but rather a sensitivity analysis conducted by comparing multiple regulatory scenarios (continued lenient operation / sudden regulatory tightening). The temporary suspension of Anthropic in June demonstrates a real-world example where existing legal frameworks like export restrictions have had an impact on AI companies that is virtually equivalent to disrupting their supply chains.

In investments in AI companies, and in the medium-term management plans of AI companies themselves, the risk that the regulatory environment itself is unstable is becoming an undeniable variable. The identity of the next official CAISI Director, and how the industry reacts to that selection, will continue to be a key indicator of the stability of future regulatory enforcement and is therefore worth watching closely.

References: www.cnbc.com / techcrunch.com / www.axios.com / thehill.com / qz.com

CAISIAI政策規制リスクファイナンス政策AI/ML

"You can apply even if you can't code"—Anthropic's $150 million investment in an unconventional talent development program for an AI company.