OpenAI had one of its busiest weeks in months: a new model family that solved decade-old math problems, a rebuilt voice stack, a public spat with Apple, and a 100,000-researcher giveaway. Meanwhile, Palantir posted a record quarter and its CEO called the AI industry Marxist, while Chinese labs kept pushing prices down. Both things are true. Progress is messy.

OpenAI’s Big Week: Astra, GPT-Live, and a Lawsuit Rebuttal

OpenAI Astra solves 10 open math problems for $2,000

On August 1, OpenAI published ten new results in mathematics and theoretical computer science, crediting the core arguments to an internal version of Astra, its next major model family. Every proof ships with a machine-checkable Lean 4 certificate, and the total inference cost was roughly $2,000 at Sol API rates. Astra itself remains unreleased. The problems had been open for at least a decade. This is not a demo. This is a signal about what the next generation of models can do when pointed at hard, verifiable problems.

GPT-Live: continuous voice interaction without turn-taking

OpenAI published a technical write-up on August 3 explaining GPT-Live, the rebuilt voice architecture now powering ChatGPT Voice. The new system lets ChatGPT listen while it speaks, dropping voice session startup from six network round trips to one. Audio moves through a dedicated fast path while reasoning happens in parallel. The result is a conversation that feels like talking to a person, not a walkie-talkie.

Apple is getting this wrong

OpenAI published a blog post on August 4 titled “Apple is getting this wrong,” rebutting Apple’s trade secrets lawsuit. The post includes private emails and messages showing that Apple’s outside lawyers emailed the wrong person after confusing two Asian last names. Business Insider called it a “brutal critique.” Fortune reported that Sam Altman’s OpenAI accused Apple of presenting a “misleading account” of the events behind the lawsuit.

GPT-5.6 efficiency gains and 100,000 academic researchers

OpenAI also published a post on GPT-5.6’s efficiency improvements: 20% lower serving costs from production GPU kernel improvements and 15%+ better token-generation efficiency from improved speculative decoding. Separately, the ChatGPT for Academic Researchers program (announced July 29) will give 100,000 researchers free access to GPT-5.6 Sol Pro through 2027, starting with 10,000 this summer.

Palantir’s Record Quarter and the Marxist Accusation

Palantir posted its best quarter ever in Q2 2026, with revenue nearly doubling to $1.9 billion from $982 million a year ago. The stock surged 29.5% intraday on August 4. CEO Alex Karp used the victory lap to torch the frontier AI labs, calling them “Marxist” in the shareholder letter and warning that they want to seize their customers’ means of production. TechCrunch reported Karp said frontier labs are “too untrustworthy for enterprises.” The irony of a defense contractor worth $150 billion calling anyone Marxist was not lost on TNW, which ran a piece titled “The one-directional radicalism of Alex Karp.”

Chinese AI Pricing War Heats Up

Alibaba released Qwen3.8-Max, a 2.4 trillion parameter model that activates about 95 billion per query, with open weights and a 1M-token context window. The stock lifted 7% on the news. DeepSeek followed with V4-Flash, which undercuts Anthropic’s Claude Fable 5 by more than 100 times on cost. Yellow.com reported that Chinese open-weight rivals are capturing a fast-growing share of global AI traffic. This is an infrastructure play, not a feature play. The pricing pressure is structural.

Canva Outdrew Every AI Chatbot But ChatGPT

Canva attracted 10.5 billion web visits between May 2025 and April 2026, ranking second among all AI platforms and trailing only ChatGPT (64.7 billion visits, or 44.76% of the entire AI tools market). The data comes from the OneLittleWeb AI Web Traffic Report covering 9,531 AI tools. Canva finished ahead of Gemini, Claude, and Grok without being a chatbot. It is a design tool that happens to have AI features baked in. That tells you something about where the real usage is.

Anthropic Hires Former California Supreme Court Justice

Anthropic named Mariano-Florentino (Tino) Cuellar as its first Chief Global Affairs Officer on August 4. Cuellar, a former California Supreme Court justice, will lead the company’s work on policy, strategic international engagement, and government relationships. CNBC reported he is tasked with finding common ground with the Trump administration, which blacklisted and ordered controls around Anthropic’s models this year. This is a serious hire for a company that needs serious diplomacy.

Quick Hits

Hugging Face released LFM2.5-2.6B, a 2.6 billion parameter model from Liquid AI designed for on-device agent deployment. It supports tool calling and multi-step workflows with a 128K context window, running on laptops and phones with no cloud dependency. Published August 4.

OpenAI Economic Research Exchange launched to study AI’s impact on jobs, productivity, and the economy. External researchers get governed access to OpenAI data. Applications are open.

OpenAI Scholars program announced alongside the Economic Research Exchange, expanding the company’s academic outreach.

Rundown for August 5, 2026. Sources: Yellow, OpenAI, Anthropic, Hugging Face, TechCrunch, Fortune, Business Insider, CNBC, TNW, DataCamp.