GPT-5 Is More Product Than Breakthrough… Maybe That’s the Point
This is the week of releases. Google drops a new world model, Genie; OpenAI rolls out GPT-5. ElevenLabs has a new music model.
GPT-5 landed this week and, on paper, it’s not the grand leap some people were expecting.
It beat benchmarks, but only by a little. And analysts reckon most of that bump is likely more from bigger context windows and better reasoning orchestration, than from a deep architectural breakthrough.
But that’s missing the plot.
Breakthroughs in AI might be slowing (invention is naturally elliptical). The biggest leaps right now might be in making AI usable.
Most of what shipped in GPT-5 was about UX, what I consider the “productization” of the model:
One unified model. GPT-5 is now the default in ChatGPT. No more picking between GPT-4o, o3, o4-mini, 4.1, 4.5. The system invisibly routes between models under the hood.
Fewer hallucinations, more transparency: It’s better at saying “I can’t do that” instead of bluffing.
Better design sense. Especially in front-end code. Layouts feel more aesthetic. We might finally be jumping from “ugly but functional” MVPs.
API developers get two new dials: verbosity lets you set length of answers without rewriting prompts and reasoning_effort lets you easily flip between “think fast” and “think deep”.
Prototypes with GPT-5 made by Pietro Schriano
This direction feels less like a moonshot, and more business-driven. Less experimental model, and more dependable software. Are we entering the productization era of OpenAI?
The shift makes sense: they have big promises to keep.
OpenAI is reportedly in talks for a secondary share sale that could value the company at $500 billion, up 1.6x in five months and 3x in the last year.
The good news: User growth is staggering with 700 million weekly actives, 4x more than last year. ARR hit $10 billion, up ~80% year-over-year.
But for all that scale, the company is still losing an estimated $5 billion a year, and projecting no cash-flow positivity until 2029 when it hits $125 billion ARR.
When the stakes look like that, it makes sense to ship the thing that makes your distribution moat deeper and your infra more efficient.
We might see more modelmakers head down this path as we search for the next exponential technical breakthrough.
Google’s new AI model creates video game worlds in real time: Google DeepMind is releasing a new version of its AI “world” model, called Genie 3, capable of generating 3D environments that users and AI agents can interact with in real time. Users will be able to generate worlds with a prompt that supports a “few” minutes of continuous interaction, which is up from the 10–20 seconds of interaction possible with Genie 2.
OpenAI launches two ‘open’ AI reasoning models: OpenAI launched two open-weight AI reasoning models with similar capabilities to its o-series. The models come in two sizes: a larger and more capable gpt-oss-120b model that can run on a single Nvidia GPU, and a lighter-weight gpt-oss-20b model that can run on a consumer laptop with 16GB of memory.
New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality: A new study from the Anthropic Fellows Program reveals a technique to identify, monitor and control character traits in LLMs. The findings show that models can develop undesirable personalities (e.g., becoming malicious, excessively agreeable, or prone to making things up) either in response to user prompts or as an unintended consequence of training. The researchers introduce “persona vectors,” which are directions in a model’s internal activation space that correspond to specific personality traits, providing a toolkit for developers to manage the behavior of their AI assistants better.
OpenAI in talks with investors about share sale at $500 billion valuation: OpenAI is in talks with investors about a potential stock sale at a valuation of roughly $500 billion. The talks are in early stages and would involve a secondary sale with shares sold by current and former employees. Thrive Capital, an investor in OpenAI, could lead the potential round, the sources said.
Perplexity accused of scraping websites that explicitly blocked AI scraping: Perplexity is crawling and scraping content from websites that have explicitly indicated they don’t want to be scraped, according to internet infrastructure provider Cloudflare. Cloudflare published research saying it observed the AI startup ignore blocks and hide its crawling and scraping activities. The network infrastructure giant accused Perplexity of obscuring its identity when trying to scrape web pages “in an attempt to circumvent the website’s preferences,” Cloudflare’s researchers wrote.
Google’s AI coding agent Jules is now out of beta: Google launched its AI coding agent, Jules, out of beta, just over two months after its public preview debut in May. Powered by Gemini 2.5 Pro, Jules is an asynchronous, agent-based coding tool that integrates with GitHub, clones codebases into Google Cloud virtual machines, and uses AI to fix or update code while developers focus on other tasks.
OpenAI is giving ChatGPT to the government for $1: OpenAI announced it will offer its ChatGPT Enterprise product to U.S. federal agencies for $1 through the next year, making its technology available to the federal executive branch workforce at “essentially no cost.”.
ElevenLabs launches an AI music generator, which it claims is cleared for commercial use: ElevenLabs announced a new model that allows users to generate music, which it claims is cleared for commercial use. This move marks ElevenLabs’ expansion beyond its main focus thus far in its three years of existence, which has been building AI audio tools. ElevenLabs also announced deals with Merlin Network and Kobalt Music Group, two digital publishing platforms for independent musicians, to use their materials for AI training.
Gemini 2.5 Deep Think is Google’s most advanced AI model to date: Deep Think, a multi-agent AI model from the Google DeepMind team, is now available via the Gemini app. The strength of Deep Think lies in its multi-agent architecture. By deploying several reasoning agents in parallel, Deep Think can process multiple ideas and concepts simultaneously. It then selects, refines, or merges the most promising threads into a final response.
Anthropic’s new Claude 4.1 dominates coding tests days before GPT-5 arrives: The new Claude Opus 4.1 model scored 74.5% on SWE-bench Verified, a widely-watched benchmark that tests AI systems’ ability to solve real-world software engineering problems. The performance surpasses OpenAI’s o3 model at 69.1% and Google’s Gemini 2.5 Pro at 67.2%.
Apple might be building its own AI ‘answer engine’: Apple has formed a new team to build a ChatGPT-like app, according to Bloomberg’s Mark Gurman. This team (reportedly called Answers, Knowledge, and Information) is working to build an “answer engine” that can respond to questions using information from across the web.
You’ve heard of AI ‘Deep Research’ tools…now Manus is launching ‘Wide Research’ that spins up 100+ agents to scour the web for you: AI startup Manus announced Wide Research, a new experimental feature that enables users to execute large-scale, high-volume tasks by leveraging parallelized AI agents, even more than 100 at a single time, all focused on completing a single task (or series of sub-tasks laddering up said overarching goal).
China debuts brain-like computer with 2 billion artificial neurons: Engineers at Zhejiang University in China have introduced a brain-inspired computer called Darwin Monkey, featuring over 2 billion artificial neurons. According to the university, the system consists of 960 Darwin 3 neuromorphic chips, creating more than 100 billion synapses. The neuron count is comparable to that of a macaque and is intended to advance brain-inspired AI research. The Darwin Monkey has reportedly completed tasks such as content generation, logical reasoning, and mathematics, using a large AI model from Chinese company DeepSeek.
Enterprises prefer Anthropic’s AI models over anyone else’s, including OpenAI’s: Anthropic now holds 32% of the enterprise large language model market share by usage, according to a report from Menlo Ventures. OpenAI holds the second-largest market share by usage among enterprises, with 25%. The figure marks a strong reversal from even just two years ago when OpenAI held 50% of the enterprise market share by usage while Anthropic had 12%.
Spending on AI data centers is so massive that it’s taken a bigger chunk of GDP growth than shopping – and it could crash the American economy (Fortune)
Demis Hassabis on our AI future: ‘It’ll be 10 times bigger than the Industrial Revolution – and maybe 10 times faster’ (The Guardian)
Windows 2030: Microsoft predicts agentic AI OS will replace mouse and keyboard (TechSpot)
Tim Cook says Apple ‘must’ figure out AI and ‘will make the investment to do it’ (The Verge)
Google denies AI search features are killing website traffic (TechCrunch)
Scientists just launched the first quantum computer into space (Futurism)
Meet Hannah Cairo. The homeschooled, self-taught 17 year old prodigy who cracked a math problem that stumped experts for 40 years.