Codex mobile/automation; enterprise tools; superapp pivot; DeployCo; Ona acquisition; Oracle deal; Partner Network; L'Oréal alliance; Jalapeño chip; BBVA partnership; Matt Wood hire; Special Situations chip deal role; Armenia education deal; HP Frontier deployment; Codex enterprise work platform plugins; AWS FDE unit competitive pressure; Workspace agents GA; app templates; Codex goal mode; free period extension; embedding engineers trend; Agent RFT fine-tuning platform; Codex reasoning token clustering bug; Sol Ultra in Codex with subagents; guardrails SDK
Key Questions
What is Codex GA and its new capabilities?
Codex has reached GA with Dell on-prem support, mobile control, and Windows computer use, now serving 5M weekly users and evolving into an enterprise work platform with role-specific plugins.
What major enterprise deals has OpenAI announced?
OpenAI has secured a $300B Oracle deal, L'Oréal alliance, Nykaa partnership, and Armenia's rollout of 50k ChatGPT Edu and Codex subscriptions, alongside HP's Frontier deployment.
What is the Jalapeño chip and its benefits?
OpenAI and Broadcom's custom Jalapeño chip was built in 9 months, matches Blackwell/TPU performance, and delivers 50% cost efficiency gains for enterprise adoption with partners like BBVA and Visa.
What is Agent RFT and how does it help enterprises?
Agent RFT is a reinforcement learning-based fine-tuning platform for enterprise agents that solves credit assignment in context windows, improving agent reliability.
What updates are in ChatGPT Enterprise and Edu release notes?
Workspace agents are now generally available with admin controls, app templates for GitHub/Snowflake/Databricks, Codex goal mode, and an extended free period to July 6.
What bug affects Codex reasoning tokens?
A clustering bug at exact token counts (516/1034/1552) impacts GPT-5.5 in 82% of cases, causing wrong results about 40% of the time and prompting community workarounds.
What guardrails were added to the Agents SDK?
OpenAI shipped guardrails in the Agents SDK, providing developers concrete safety architecture for building reliable agents.
How are companies managing API usage limits?
Practical guides now cover setting OpenAI API usage limits to prevent agent overspending, with admin tools available in Enterprise and Edu workspaces for monitoring overages.
Codex GA with Dell on-prem, mobile control, Windows computer use; 5M weekly users, 99.8% of internal AI tokens; DigitalOcean plugin for remote sessions. Codex evolving into a full enterprise work platform with role-specific plugins for investment banking, sales, and data analysis. HP deploying OpenAI Frontier to scale enterprise AI with concrete metrics (122 PRs, one-day bug fix). DeployCo $4B consulting subsidiary. Frontier enterprise agent platform with BCG. OpenAI acquires Ona. $300B Oracle deal. Partner Network with $150M backing. L'Oréal alliance, Nykaa partnership. OpenAI and Anthropic poaching enterprise sales leaders from Salesforce. OpenAI and Broadcom announce custom chip 'Jalapeño' built in 9 months, on par with Blackwell/TPU, 50% cost efficiency gain; enterprise adoption with BBVA and Visa. Matt Wood hired to expand cloud partnerships beyond Azure. 'Special Situations' role created for semiconductor partnerships. Armenia rolling out 50k ChatGPT Edu and Codex subscriptions as education push. Samsung enterprise deployment noted as new adoption signal. OpenAI models now FedRAMP High and DoD IL-4/5 approved on Bedrock via Mantle. ChatGPT Webflow integration now available, reducing copy-paste friction for website management. AWS launches $1B Forward Deployed Engineering unit, intensifying enterprise AI deployment competition against OpenAI's DeployCo and Microsoft's Frontier Co. New: ChatGPT Enterprise & Edu release notes: Workspace agents GA with admin controls, app templates for GitHub/Snowflake/Databricks, Codex goal mode, free period extended to July 6. New: A massive shift from selling AI models to embedding engineers on-site is underway, with $9B collective bet across OpenAI, Anthropic, Amazon, Meta, and Microsoft; MIT study shows 95% of pilots fail, reinforcing the need for deployment services. New: OpenAI launches Agent RFT, a reinforcement learning-based fine-tuning platform for enterprise agents, solving credit assignment in context windows. New: Practical guide on setting OpenAI API usage limits to stop agent overspending now available. New: Codex reasoning token clustering bug at exact counts (516/1034/1552) with GPT-5.5 accounting for 82% of exact-516 events, causing wrong results ~40% of time; community workarounds and frustration signal reliability concern for enterprise users. New: GPT-5.6 Sol Ultra confirmed heading to Codex with subagent cooperation, making Ultra more viable for enterprise coding workflows. New: OpenAI shipped guardrails in their Agents SDK — a concrete product update for developers building agents, revealing internal safety architecture.