Jensen Huang's first post on X was a three-page PDF, and the paragraph in it that carries the actual ask is the ninth, not the thesis line that got quoted everywhere. A day later Dharmendra Pradhan posted his resignation as two images of Hindi text, and the same construction sits inside it: one specific failure conceded in a passive clause, then the reason for quitting placed on other people. Both letters are reproduced below at length, because in both cases the argument is in the sentences nobody pulled out.
24 July 2026
Source: "Open Weights and American AI Leadership", three pages, dated 24 July 2026, hosted at images.nvidia.com and mirrored by Microsoft, attached as three photos to Jensen Huang's debut post on X.
"In the 1980s, early open-source software pioneers challenged the prevailing belief that software would advance only if companies kept tight control over their code. This movement pushed for a transparent ecosystem where developers around the world could study, modify, and improve software. Software developed by the open-source community now supports most of the internet and underlies systems used by the world's largest technology companies, as well as the U.S. military and federal agencies conducting scientific research, cybersecurity, and other critical missions. Open source did more than lower the cost of software; it created a shared foundation of knowledge on which generations of American engineers and entrepreneurs built their institutional sovereignty."
"The United States now faces a similar choice with artificial intelligence. Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector. This is essential for creating opportunities for innovation and prosperity across the country. It requires expanding access to AI, encouraging competition, robust application layers, and giving Americans greater control over the technology they rely on. Open-weight models, AI models that anyone can download, inspect, modify, and run on their own infrastructure, are an important part of that foundation because they make advanced AI more accessible, adaptable, and widely available."
"Open weights let every organization match the right model to the right job at the right cost, reserving frontier-scale capability for genuine frontier problems and running efficient, specialized models everywhere else. That discipline is what will make AI economically sustainable as its use scales into the billions of everyday tasks. America wins the AI era by diffusing it into the workflows of factories, hospitals, farms, classrooms, and main street businesses."
"To be sure, open weights carry real and distinct risks. Once released, the weights are beyond the original developer's control, and modified versions are difficult to trace or reverse. But the right response to this risk is not to prohibit open weights. In a world where cybersecurity attackers use advanced AI, defenders need access to models with comparable capabilities so they can detect, simulate, and respond to emerging threats."
"In fact, openness may be one of the most important paths to AI safety and security. Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect. And concentrating advanced AI capabilities behind a small number of closed models compounds that risk. It results in a small number of single points of failure, weakens competition, and leaves critical technology in the hands of a few providers."
"A strong AI ecosystem is not a foregone conclusion. Policymakers have an important opportunity to act. This includes expanding access to compute for startups and researchers, investing in shared training assets (datasets, tools, evaluation frameworks), and keeping the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas."
"In shaping this ecosystem, policymakers should be careful not to conflate legitimate model-development techniques with misappropriation. Distillation, or the practice of using one model's outputs to help train or improve another, is a widely used technique for model improvement, evaluation, and validation. It reflects a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement. By contrast, unlawful efforts to extract value from closed models raise legitimate concerns. Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation."
"American Innovators Network • Andreessen Horowitz • Arcee AI • Arena • Black Forest Labs • Box • CrowdStrike • Dell Technologies • Emergence Capital • Hugging Face • IBM • The Linux Foundation • Mariana Minerals • Meta • Microsoft • Mistral • Mozilla • NVIDIA • Palantir • Perplexity • Reflection • Replit • ServiceNow • Telnyx • Y Combinator"
The word China does not appear anywhere in those three pages, and no model or incident is named. Two days before publication, OSTP Director Michael Kratsios had accused Moonshot AI of running large-scale distillation against U.S. models to build Kimi K3, released on 16 July 2026 as a 2.8T-parameter open-weight model, and Treasury Secretary Bessent had said sanctions were on the table. The paragraph about distillation is the response to that, written as a general principle about technique.
Endorsement by reply
Source: X, 24 and 25 July 2026.
Huang, @JensenHuang: "For my first post, I'm sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models."
Sundar Pichai, @sundarpichai: "Very happy to support this on behalf of Google. We have long benefited from open source, are big contributors to open source and in fact have consistently made open weights models with Gemma available from @GoogleDeepMind @demishassabis . Onwards!"
Julian Schrittwieser, @Mononofu, Anthropic technical staff: "I'm so excited that @JensenHuang is a believer in open source now, looking forward to the CUDA and GPU driver open source release!"
A reply to Schrittwieser: "Nobody is saying that all software has to be open source. What they're saying is that open-weight AI should be allowed. By implication, they're rejecting your company's incessant machinations to kneecap the open model ecosystem. Read the room."
A reply in the Sacks thread: "Not just this, but think about the premise: Anthropic's models are trained on illegally torrented, pirated, like, 'grab everything from github, PhD papers, pirate each book, then train on it'. It's really a non-sequitur to even make the analogy. CUDA never stole anything."
Pichai's line went out while Google's logo was not on the sheet above, and so did Demis Hassabis's, who cited more than 300 million Gemma downloads. That is a CEO endorsing the substance faster than his own legal and comms process could put a logo in a PDF. It reads as more than posture to me: Gemma 4(2 April 2026) ships under Apache 2.0 where Gemma 3 and earlier were source-available under Google's custom terms, which is exactly the change that used to make enterprise legal review painful. Gemini stays closed, so it is a two-track strategy, not a conversion. The signature count went from 25 at launch to about 50 within a day, with OpenAI and Google both joining, and Microsoft's page showed 270+ by 3 August. Anthropic and Amazon stayed off it.
"CUDA never stole anything"
The CUDA jab is a category error and the reply to it is right about that much: asking that open weights remain legal is a claim about permission, not an obligation on every vendor to open every product. Schrittwieser clarified the next day that he thinks "open models can be very useful" and that his point was about "historically extremely anti-open source companies" suddenly favouring openness, which is a fair jab, and Amodei's post of 27 July says flatly that "Anthropic has never advocated for a ban on open-weights models." His actual asks are chip export controls, a crackdown on industrial-scale distillation, and mandatory pre-release safety testing for sufficiently capable models. The third one is the real fight, since a compliance floor only well-funded labs can clear does the work of a ban without the word.
The piracy reply is half solid and half invented. Judge Alsup found on 23 June 2025 in Bartz v. Anthropic that the company downloaded over seven million books it knew were pirated from Books3, LibGen and PiLiMi, that training on them was fair use, and that acquiring and keeping the pirated copies was not. The settlement is $1.5B, roughly $3,000 per work across about 500,000 books, given final approval on 20 July 2026. That part is on the record. GitHub is NOT: no adjudicated finding says Anthropic torrented repositories, the code-training case is Doe v. GitHub against other companies, and the Bartz class was defined by ISBN and ASIN, so "PhD papers" is not in it either. And "CUDA never stole anything" skips that NVIDIA itself was sued in March 2024 over training NeMo Megatron on Books3, the same pirated corpus. Both sides here are arguing their own income statement. NVIDIA, Dell, IBM and the clouds sell more when models are cheap and interchangeable, Anthropic sells scarcity, and the open ecosystem the letter says America must lead is currently Chinese by download volume: the ATOM report puts Qwen at 942.1M Hugging Face downloads against roughly 56M for all U.S. open entrants combined. On 5 August, according to reporting, the administration told AI companies the new security review framework would apply to closed frontier models only, with open-weight models exempt. The letter's side won the round in twelve days.
25 July 2026
Source: two images posted by @dpradhanbjp, 25 July 2026, the resignation letter of the Union Education Minister, in Hindi, addressed "मेरे युवा साथियों". Translation below is mine, the original wording is in the image.
"My young friends,"
"For more than the last 4 decades I have been dedicated to students, teachers and education reform. It has always been my belief that a strong, inclusive and far-sighted education system is the very foundation of a strong nation."
"I deeply respect the aspirations, the feelings and the just expectations of the country's youth. Realising the dreams of India's young generation has been the moral resolve of our political and social life. I express my gratitude to the Prime Minister for the opportunity I received to serve the nation under his visionary leadership."
"But irregularity was found in the NEET-UG examination held on 3 May 2026. The Government of India, taking immediate cognisance of this, handed the investigation to the CBI and cancelled the examination, and the date of the re-examination was announced. Along with this, the decision was taken to conduct this examination in CBT (Computer-based-test) mode from next year."
"During this period our chief priority was that the examination of more than 20 lakh students be conducted smoothly. For this, work was done under a whole of government approach. In this, along with the Government of India, the state governments and especially the district administration played an important role. With the cooperation of students, guardians and parents, this examination was completed on 21 June 2026."
"From the very first day I took responsibility for this and never turned my face away from the situation. My resolve was that we would not let the prospects of any meritorious student be ruined by the exam mafia, that injustice would not be done to any student."
"The NEET-UG results declared on 16 July were satisfactory, in which many meritorious students from poor backgrounds also found success."
"But even during this period, persons sitting in positions of responsibility tried to obstruct, in order to mislead many students, which distressed my mind greatly."
"I have always held unbroken faith in the power of our democracy and have always deeply respected the aspirations, dreams and expectations of the youth. They are not only India's future, but the carriers, builders and architects of a new and developed India."
"Seeing the events of the last 10 days my mind is saddened. This is not a matter of personal prestige for me."
"India's youth power is the real strength of this country."
"We will not let the youth power of the country be trapped in a vicious cycle of confusion, this is my resolve."
"Considering the situation that has arisen at Jantar Mantar and in the country, so that anti-national forces do not take advantage of this situation, so that the unity of the country remains, so that the future of not one student of India gets entangled in legal complications, so that our children put their time into study and focus on building a career, thinking on exactly this I have sent my resignation to the Hon'ble Prime Minister."
(A paragraph of thanks to the Prime Minister, to cabinet colleagues and to ministry officials follows.)
"With the blessings of Lord Shri Jagannath, in future too I will keep myself dedicated in every possible way to Bharat Mata, to the people of Odisha, and to fulfilling the aspirations of the country's youth."
"Yours, Dharmendra Pradhan"
After the resignation
"Irregularity was found" is the whole admission, and it has no subject. What the CBI chargesheet reporting describes is not an irregularity that was found, it is a leak run by three NTA subject experts who had selected students write the questions down in coaching sessions, which is why nothing circulated digitally until an accused reconstructed the paper and left a trail. Rajasthan police got the tip-off on 7 May, the recovered document matched 120 questions including all 90 Biology and all 45 Chemistry, per ADGP Vishal Bansal, and the exam was cancelled on 12 May. The insider route is the same failure the K. Radhakrishnan panel was set up to prevent after 2024, so this is a repeat, and no letter that puts the blame on "persons sitting in positions of responsibility" outside the ministry survives that chargesheet. The "whole of government approach" sentence also sits oddly next to the retest of 21 June, where roughly 700,000 personnel were deployed, Telegram was blocked nationally until 22 June, and reportedly more than 2.5 lakh candidates did not turn up at all. Results on 16 July: 11.21 lakh qualified, over 58% of them women.
On the Jantar Mantar line, I do not think it is the absurd part. Cockroach Janta Party was founded on 16 May 2026 by Abhijeet Dipke, 30, a political communications strategist and Boston University graduate who had earlier worked with AAP, and it passed 20 million Instagram followers, more than the BJP or the Congress hold. That is a professionally run campaign, not a spontaneous assembly of aspirants, and the government treating it as one is not paranoia. What undercuts the letter is that the same government conceded all three of the CJP demands the same evening, compensation for families, withdrawal of cases against protesters, and the minister's own exit, which is a settlement, not a defence against anti-national forces. As such, the letter reads as the exit note of a minister who was removed by the arithmetic of ~2 million exam-takers and their parents, and who wrote around it.
Then the one testable claim in either document. Pradhan writes that the decision was taken to run NEET-UG in CBT mode from next year, and NTA told the Supreme Court on 29 May the same thing. In an August 2026 affidavit the Centre told the Court that migration to CBT is only "under active consideration", pending an expert task force report and stakeholder consultation. Parliament has since passed the Public Examinations amendment with two-month investigation and three-month trial deadlines that PRS notes are routinely missed under comparable statutes, and the NTA overhaul announced around 18 August removed 600 experts and moved offices into CISF-secured premises. All of that is downstream. One notification from Pralhad Joshi's ministry naming the mode for NEET-UG 2027 and the date that decision was actually taken would close the gap between the letter and the affidavit, and it has been available to publish since August.