OpenAI Reportedly Preparing IPO Filing Within Days to Weeks

According to the Wall Street Journal, OpenAI is preparing to file for an initial public offering (IPO) "within days to weeks." Banks including Goldman Sachs and Morgan Stanley are helping to prepare the prospectus, and the filing could come as soon as Friday.

OpenAI declined to elaborate, saying in a statement only that it "regularly evaluates strategic options." But an IPO of the AI industry's largest startup would undoubtedly have a major impact on the market as a whole. Its latest reported valuation already stands at around $300 billion, which could make it a historically large technology IPO.

AI Chatbots Fail 90% of the Time on Election News Accuracy

A new study by Forum AI found that the four major chatbots — ChatGPT, Google Gemini, Anthropic Claude, and xAI Grok — failed 90% of the time on at least one of "accuracy, bias, or source selection" in their responses to questions about elections and geopolitics.

After more than 3,100 questions were asked, election-related responses in particular showed significant inaccuracies. With the US midterm elections just around the corner, the unreliability of AI chatbots as an information source is a serious concern. Improving fact-checking accuracy has become an urgent priority for AI companies.

Google AI Studio Generates Native Android Apps from Prompts

Google AI Studio can now generate native Android apps from prompts. The apps are built with Kotlin and Jetpack Compose and can be tested in a browser-based emulator.

The Decoder describes this as "the app market version of the SaaSpocalypse." For simple utility apps such as trackers and checklists, the Play Store's raison d'être could fade. Apple, meanwhile, has consistently blocked vibe-coding-style apps, making the contrast between the two companies' approaches particularly interesting.

While AI-powered code generation accelerates the democratization of app development, it will also give rise to challenges in quality assurance and security.

Airbnb CEO Defends Use of Chinese AI Model (Qwen)

Airbnb CEO Brian Chesky pushed back against a US House investigation into the company's use of Alibaba's Qwen large language model in its customer support chatbot, arguing that "US lawmakers misunderstand the technology."

"We don't give data to Chinese companies. They don't have access to any data," he said, emphasizing that using an open-source model does not equate to a data leak. These remarks add fuel to the debate over the use of open-source AI models in the context of US-China tensions.

As Chinese open-source models continue to improve in performance, how to balance security and capability is becoming a decision many companies now face.

Tool Calling Cost Comparison: Real-World Results Show 2% Quality Gap at 10x the Cost

A detailed cost comparison of tool calling using five models (Opus 4.7, GPT 5, Sonnet 4.6, DeepSeek V4 Pro, and Tencent Hunyuan) is drawing attention on Reddit's MachineLearning community.

After running 8 refactoring tasks on a 15k-line Python project, the quality gap between the most expensive model, Opus 4.7 (~$15), and the cheapest, Tencent Hunyuan (~$1.50), was less than 2 percentage points. It turns out that for everyday refactoring, cheaper models work perfectly well.

That said, expensive models were overwhelmingly superior at refactoring complex nested decorators, underscoring the importance of routing based on task complexity. Interestingly, the poster also ran local inference on an M2 Ultra Mac Studio (192GB), cutting daily spending from around $40 to $9.

---

**References:**

[OpenAI preparing for IPO filing in days or weeks, WSJ reports](https://www.bloomberg.com/news/articles/2026-05-20/openai-preparing-for-ipo-filing-in-days-or-weeks-wsj-reports?srnd=phx-technology) — Bloomberg

[Chatbots Struggle With News Accuracy and Sourcing Ahead of Midterms](https://www.bloomberg.com/news/articles/2026-05-20/chatbots-struggle-with-news-accuracy-sourcing-ahead-of-midterms?srnd=phx-technology) — Bloomberg

[Google tests the app market version of the SaaSpocalypse](https://the-decoder.com/google-tests-the-app-version-of-the-saaspocalypse/) — The Decoder

[Airbnb CEO says US misunderstanding use of Chinese open-source AI models](https://www.bloomberg.com/news/articles/2026-05-20/airbnb-s-chesky-says-us-misunderstanding-use-of-chinese-open-source-ai-models?srnd=phx-technology) — Bloomberg

[under 2% quality gap but 10x cost difference: tested 5 models on identical tool calling tasks](https://www.reddit.com/r/MachineLearning/comments/1tiqsao/under_2_quality_gap_but_10x_cost_difference/) — Reddit MachineLearning