How independent these versions are. Agreement between versions is evidence only where they were compiled independently. Claude wrote first and drafted the brief. ChatGPT read Claude's draft before compiling, so overlap between those two is not independent agreement. Grok compiled its selection for this week independently. DeepSeek compiled its selection without seeing the others but could not browse; its sources were supplied and confirmed by the collection.
Stories across the versions:
| Story | Selected by |
|---|---|
| DeepSeek-R1 | Claude, ChatGPT, DeepSeek, Grok |
| Stargate | Claude, ChatGPT, DeepSeek, Grok |
| Operator | Claude, ChatGPT, DeepSeek, Grok |
| Trump revokes Biden's order; orders an action plan | Claude, ChatGPT, DeepSeek |
| Meta's $60–65B capex plan | Claude, ChatGPT |
| Gemini 2.0 Flash Thinking update | Grok |
Claude (Claude Opus 5.5)
1. DeepSeek Releases R1, an Open Reasoning Model Matching o1
Date: January 20, 2025
Context: DeepSeek released R1 under the MIT license, together with six smaller distilled models and a technical report crediting large-scale reinforcement learning. The company said R1 performs on par with OpenAI's o1 in math, coding and reasoning. Its API was priced at a small fraction of o1's.
Significance: It was the first open-weights reasoning model claimed to match o1, released at a small fraction of o1's API price.
Feedback: In a same-day write-up, Simon Willison called the release of both large models plus six distilled versions "a fascinating flex." He noted the MIT license but questioned how it applied to the Llama-based distills, and his own tests of a small distilled model gave mixed results.
Sources:
2. Stargate: Up to $500 Billion for U.S. AI Infrastructure
Date: January 21, 2025
Context: At the White House, President Trump announced the Stargate joint venture of OpenAI, SoftBank and Oracle, with MGX as an equity funder. The partners committed $100 billion immediately and up to $500 billion over four years, with construction already under way in Texas.
Significance: It was the largest AI infrastructure commitment yet announced, and it tied OpenAI's compute plans closely to the new administration.
Feedback: On January 22, Elon Musk said the backers "don't actually have the money," prompting a public exchange with Sam Altman.
Sources:
3. Trump Revokes Biden's AI Order and Calls for a New AI Action Plan
Date: January 20 and 23, 2025
Context: On his first day in office, President Trump revoked Executive Order 14110, Biden's 2023 framework for AI safety. On January 23 he signed "Removing Barriers to American Leadership in Artificial Intelligence," which ordered an AI action plan within 180 days.
Significance: Federal AI policy turned from managing risk to promoting competitiveness.
Feedback: No independently sourced reaction was located within the week.
Sources:
Hindsight: on July 23, 2025, the White House released "America's AI Action Plan." Source: White House[link]; see also Akin Gump[link]. Cross-reference: the series' post for July 21–27, 2025.
4. OpenAI Launches Operator, Its First Computer-Using Agent
Date: January 23, 2025
Context: OpenAI released Operator as a research preview for U.S. Pro subscribers. Operator is an agent that uses its own browser to fill out forms, order groceries and complete other web tasks. It is powered by a new "Computer-Using Agent" model.
Significance: It was the first major consumer launch of a general web agent, the concrete form of Altman's prediction that agents would join the workforce.
Feedback: No independent reaction was located within the week. At the January 23 launch demo, OpenAI researcher Yash Kumar acknowledged that Operator "still makes mistakes" (MIT Technology Review).
Sources:
Hindsight: early hands-on accounts, including Bloomberg's Rachel Metz, described Operator as slow and needing close supervision. Source: TweakTown (February 3)[link]
5. Meta Plans $60–65 Billion in 2025 Capital Spending
Date: January 24, 2025
Context: Mark Zuckerberg said Meta would spend $60–65 billion in 2025, mostly on AI. The plans included about 1.3 million GPUs by year-end and a data center drawing more than 2 gigawatts, which he said would cover a large part of Manhattan.
Significance: Three weeks after Microsoft's figure and three days after Stargate, it showed the top companies all scaling up compute at once.
Feedback: No independently sourced reaction was located within the week.
Sources:
ChatGPT (GPT-6)
1. DeepSeek releases R1 and smaller distilled reasoning models
Date: January 20, 2025 — release.
Context: DeepSeek released R1, reporting results comparable to OpenAI’s o1 on selected reasoning, mathematics and coding evaluations. It also released six smaller distilled models derived from Qwen and Llama models. The release provided downloadable weights as well as hosted access, giving developers several ways to investigate the claims.
Significance: Competitive reasoning performance was becoming accessible through more than closed hosted services. The smaller models made local experimentation possible for developers unable to run the full system.
Feedback: In his January 20 hands-on account, Simon Willison emphasized the smaller models’ accessibility and questioned how their stated licensing related to the underlying Llama license. His post was an experiment and commentary, not a comprehensive benchmark audit.
Sources:
- DeepSeek, January 20, primary release announcement: deepseek.com[link]
- Simon Willison, January 20, hands-on response: simonwillison.net[link]
2. Stargate announces a plan for up to $500 billion in US AI infrastructure
Date: January 21, 2025 — announcement.
Context: OpenAI announced Stargate, a new company intended to invest $500 billion over four years in US AI infrastructure. The announcement described an initial $100 billion deployment and named SoftBank, OpenAI, Oracle and MGX as initial equity funders. The headline described an investment ambition, not a completed expenditure or a fully demonstrated financing package.
Significance: The proposal placed AI infrastructure on the scale of major national industrial projects. Financing and execution were immediate questions alongside the demand for additional computing power.
Feedback: Elon Musk questioned the financing; Sam Altman disputed his claim. Al Jazeera reported the exchange on January 22.
Sources:
- OpenAI, January 21, primary announcement: openai.com[link]
- Al Jazeera, January 22, contemporary report of the disagreement: aljazeera.com[link]
3. Operator gives ChatGPT Pro users a browser-operating agent preview
Date: January 23, 2025 — research-preview launch.
Context: OpenAI introduced Operator for ChatGPT Pro users in the United States. It used a browser interface to carry out tasks by interacting with pages, rather than relying solely on dedicated integrations. The preview included user handoffs and confirmations, with limitations on sensitive actions.
Significance: Operator brought the question of reliable action into an everyday interface: the web browser. Successful deployment required more than understanding an instruction; it required handling changing pages, errors and moments needing human judgment.
Feedback: Simon Willison’s January 23 response compared it with Anthropic’s computer-use demonstration and highlighted OpenAI’s hosted browser as a more packaged product. He also noted the cautious confirmation and takeover behavior shown in the demonstration.
Sources:
- OpenAI, January 23 launch text; exclude later updates: openai.com[link]
- Simon Willison, January 23, contemporary analysis: simonwillison.net[link]
4. Trump orders a new AI action plan
Date: January 23, 2025 — executive order.
Context: President Trump signed an order directing officials to produce an AI action plan within 180 days. It also instructed them to review policies and actions taken under the prior administration’s AI order for consistency with the new administration’s approach. The new plan itself had not yet been produced.
Significance: The order established a change in federal policy direction while leaving substantial implementation questions open. For organizations making long-term AI investments, the distinction between a policy instruction and settled rules was consequential.
Feedback: No independently sourced reaction was located within the week.
Sources:
- White House, January 23, primary order: whitehouse.gov[link]
- Associated Press, January 23, contemporary coverage: apnews.com[link]
5. Meta sets out a $60–65 billion capital-spending plan
Date: January 24, 2025 — announcement.
Context: Mark Zuckerberg said Meta planned capital expenditures of $60–65 billion during 2025 as it expanded infrastructure for AI. He described a large increase in computing capacity, including an expectation of more than 1.3 million GPUs by year-end. These were spending and capacity targets, not completed purchases.
Significance: Coming three weeks after Microsoft’s announcement, Meta’s plan provided another measure of the resources major platforms were committing to AI. The comparison also required care: corporate capital-expenditure totals do not all use identical scopes or fiscal periods.
Feedback: No independently sourced reaction was located within the week.
Sources:
- Reuters, January 24, contemporary report: reuters.com[link]
- Data Center Dynamics, January 24, infrastructure coverage: datacenterdynamics.com[link]
DeepSeek V4.1-Flash-1017
1. DeepSeek releases open-weight R1 reasoning model
Date: Event date, January 20, 2025.
Context: Chinese startup DeepSeek released DeepSeek-R1, an open-weight reasoning model that it said was trained at a fraction of the cost of comparable US models. The release included a technical paper detailing its reinforcement-learning approach.
Significance: R1 demonstrated that a relatively small lab could produce competitive reasoning performance with far less compute, challenging assumptions about the capital requirements for frontier AI.
Feedback: No independently sourced reaction was located within the week.
Sources:
2. Trump announces Stargate, a $500 billion AI infrastructure joint venture
Date: Event date, January 21, 2025.
Context: President Donald Trump announced that OpenAI, SoftBank, and Oracle would form a joint venture called Stargate to build AI data centres in the US, with an initial $100 billion commitment and a stated target of up to $500 billion over four years. The announcement was made at the White House with the CEOs of the three companies present.
Significance: Stargate signalled a policy shift toward direct executive-branch support for large-scale private AI infrastructure, and tied the new administration's industrial policy to the ambitions of leading AI firms.
Feedback: No independently sourced reaction was located within the week. The executives' own remarks at the announcement: SoftBank's Masayoshi Son told Trump, "We wouldn't have decided unless you won," and Sam Altman said, "we wouldn't be able to do this without you, Mr. President."
Sources:
3. Trump signs executive order revoking Biden's AI safety order
Date: Event date, January 20, 2025 (revocation) and January 23, 2025 (new order).
Context: On his first day in office, Trump revoked President Biden's 2023 executive order on AI safety. On January 23 he signed Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence," directing the preparation of a new national AI action plan within 180 days.
Significance: The reversal dismantled the previous administration's formal AI-safety and reporting framework and replaced it with an innovation-and-dominance framing, immediately altering the regulatory environment for AI developers in the US.
Feedback: No independently sourced reaction was located within the week.
Sources:
4. OpenAI releases Operator, a browser-using AI agent
Date: Event date, January 23, 2025.
Context: OpenAI launched Operator as a "research preview" for ChatGPT Pro subscribers in the US. The agent, powered by a new Computer-Using Agent (CUA) model, could autonomously navigate websites to perform tasks such as filling forms, booking reservations, and shopping.
Significance: Operator was one of the first mainstream products to move AI from text generation to autonomous action in a browser.
Feedback: No independently sourced reaction was located within the week.
Sources:
5. Slot left empty.
Grok 4
1. DeepSeek releases DeepSeek-R1
Date: January 20, 2025
Context: DeepSeek released the open-weight reasoning model DeepSeek-R1 (671B-parameter MoE, ~37B active), trained largely via reinforcement learning, claiming performance comparable to OpenAI o1 on math, code, and reasoning, with weights and technical report under MIT license plus distilled variants. API and website access launched the same day.
Significance: It showed a Chinese lab matching leading closed reasoning models at lower reported cost while releasing weights openly.
Feedback: No independently sourced reaction was located within the week.
Sources:
Hindsight: Market reaction and App Store surge followed the next week (later reports: techcrunch.com[link] ; [link]).
2. Announcement of the Stargate Project
Date: January 21, 2025
Context: OpenAI, SoftBank, Oracle, and MGX announced the Stargate Project to invest up to $500 billion over four years ($100 billion immediately) in U.S. AI infrastructure. SoftBank and OpenAI were lead partners; technology partners included Arm, Microsoft, NVIDIA, and Oracle. Construction had begun in Texas. President Trump participated in the White House announcement.
Significance: It was one of the largest private AI infrastructure commitments announced to date.
Feedback: No independently sourced reaction was located within the week.
Sources:
3. Google releases updated Gemini 2.0 Flash Thinking experimental model
Date: January 21, 2025
Context: Google released gemini-2.0-flash-thinking-exp-01-21 with a 1-million-token context window, native code execution, longer outputs, and reported scores of 73.3% on AIME and 74.2% on GPQA Diamond. Available in AI Studio and the API.
Significance: It advanced Google's explicit reasoning models while remaining free/experimental.
Feedback: No independently sourced reaction was located within the week.
Sources:
4. OpenAI launches Operator research preview
Date: January 23, 2025
Context: OpenAI released Operator, a research-preview agent using a Computer-Using Agent model to perform multi-step browser tasks (reservations, shopping, forms). Initially available to U.S. ChatGPT Pro subscribers.
Significance: It was OpenAI's first public general-purpose web-acting agent.
Feedback: No independently sourced reaction was located within the week.
Sources: