OpenAI GPT-5.5 Instant/Pro/Codex rollout
Key Questions
What new features are included in the GPT-5.5 Instant rollout?
GPT-5.5 Instant is now the default model and includes memory integration, Gmail access, and Codex capabilities along with Realtime-2 and Images 2.0. It also introduces six job-specific Codex plug-ins and a Sites feature.
How much productivity gain has Codex demonstrated in case studies?
A Codex case study showed a reduction from 180 hours to 15 hours in task completion time. OpenAI claims GPT-5.5 is currently the best coding model available.
What does the MIT study reveal about AI coding tool productivity?
The MIT study found that AI coding tools increase code volume by 300% but only boost actual output by 30%, challenging common productivity claims. It highlights that adoption friction is often organizational rather than related to model quality.
How is Codex expanding beyond traditional coding tasks?
Codex is being used in non-coding domains such as running farms, building iOS apps, and working with Figma. Adoption in India has increased 27x since January, with 5 million weekly active users overall.
What is OpenAI's plan for turning ChatGPT into a super app?
OpenAI intends to expand ChatGPT into a super app with broader capabilities before going public. This includes deeper integration of tools like Codex for knowledge workers, who are growing 3x faster than developers.
What concerns exist around GPT-Rosalind?
GPT-Rosalind is a hidden biology model from OpenAI that has raised dual-use concerns due to its potential applications. Details remain limited in public announcements.
What is the FrontierCode benchmark?
FrontierCode is a new benchmark that measures code mergeability rather than just test-passing rates. It aims to provide a more practical evaluation of coding model performance.
How many knowledge workers are using Codex?
Codex has 5 million weekly active users, with 20% identified as knowledge workers. Their adoption rate is growing three times faster than that of developers.
GPT-5.5 Instant default + memory/Gmail/Codex; Realtime-2; Images 2.0; Codex case study shows 180h→15h productivity gain. GPT-5.5 crowned best coding model. GPT-Rosalind hidden biology model with dual-use concerns. OpenAI launches six job-specific Codex plug-ins and Sites feature; 5M weekly active users, 20% knowledge workers growing 3x faster than devs. Codex adoption in India up 27x. Codex expanding beyond coding (farms, iOS apps, Figma). OpenAI plans to turn ChatGPT into 'super app' before going public. MIT study claims AI coding tools boost code volume 300% but output only 30%, challenging productivity narrative. FrontierCode benchmark emerges, measuring mergeability over test passing. New economic perspective: adoption friction is organizational, not model quality. Hacker News discussion echoes MIT study: more AI code doesn't necessarily speed up teams.