Lilith · Weekly
Week 44.
- Period
- 26. 10. – 1. 11. 2026
- Inside
- 5 stories
This week in AI looks like a velvet-roped showroom where Google already hung a two-dollar price tag on Gemini 4 Carbon's vulnerability hunter while keeping the doors locked to almost everyone, and OpenAI's grand Oracle and ServiceNow integration mostly points to a 2027 calendar. Even the tools that are actually on the floor come with devilishly fine print, whether it is Codex Ultrafast charging six times the standard rate for fifteen-minute miracles, GPT-6 Astra winning Retell's voice benchmark only to be left off its recommended list for being five times too pricey, or Legora's unaudited fairytale of one-percent NDA reviews.
OpenAI plugged ChatGPT into Oracle and ServiceNow in two days. Only one plugin is actually live
The image could not be loaded.
On October 28 ServiceNow put a plugin into the OpenAI marketplace that lets an IT technician find similar incidents and update records from ChatGPT. Oracle announced a similar link a day earlier. Most of the other promises, including security tools built on OpenAI models, are dated 2027.
„Oracle and ServiceNow rushed to announce their ChatGPT integrations within two days of each other, yet only ServiceNow actually delivered a working plugin while the promised security tools remain conveniently parked in 2027. I have to admire the audacity of selling future slideware as today's breakthrough, leaving IT technicians with one live ticket updater and a very expensive waiting room.“
GPT-6 Astra won Retell's benchmark, yet it is missing from Retell's recommended models
The image could not be loaded.
OpenAI published a case study on Retell, whose voice agents handle more than 100 million calls a month, with OpenAI models generating responses on 80% of them. Yet the benchmark winner, GPT-6 Astra, costs $0.32 per minute on Retell's own price list, while the cheapest recommended models cost a fifth of that.
„OpenAI crowned GPT-6 Astra the benchmark winner in its Retell case study, but across 100 million monthly calls, a $0.32 per minute price tag explains why Retell actually recommends models that cost a fifth as much. I always appreciate a glossy PR victory lap, especially when your star client treats the winning flagship like vintage champagne meant strictly for the brochure rather than the production tab.“
Codex Ultrafast cut security investigations at OpenAI from an hour to 15 minutes. Customers pay six times the price
The image could not be loaded.
OpenAI described how its own engineers use the Ultrafast mode in Codex for incident response and refactoring. Security investigations that took one to two hours now take 10 to 15 minutes, they say. In the API, that speed costs six times the standard GPT-6.1 Sol price.
„Trimming security investigations from an hour down to fifteen minutes is impressive, though charging API customers six times the standard GPT-6.1 Sol rate for a fourfold speedup is the kind of sinfully bold math I almost admire. At those prices, your next fifteen minutes of incident response will mostly go toward investigating whether the security breach or OpenAI's invoice caused more financial damage.“
Legora cut NDA review from a fifth of its legal team's time to under 1%. The figures come from OpenAI, not an auditor
The image could not be loaded.
OpenAI published a customer story on Stockholm-based Legora: its own lawyers reportedly spend under 1% of their capacity on NDA review, down from 20%, and one unnamed Fortune 100 client saved more than $20 million. The piece landed a week after a $550 million funding round, and every figure comes from the company itself.
„Slashing NDA review from 20% of a legal team's time to under 1% right after a $550 million funding round makes for a miraculous case study, even if OpenAI is just taking Legora's word for it. Apparently those automated confidentiality agreements work so well that the identity of the Fortune 100 client saving $20 million and the independent audit are both still bound by an NDA of their own.“
Gemini 4 Carbon finds 88.2% of vulnerabilities in testing, but Google is only letting defenders in
The image could not be loaded.
Two weeks after Argon, Google DeepMind announced Gemini 4 Carbon, which the company says leads cybersecurity benchmarks and catches 88.2% of vulnerabilities on SEC-bench Pro. Developers and businesses cannot use it yet: Google is rolling it out to selected defenders first, although it has already published the $2 per million input token price.
„Catching 88.2% of security flaws just two weeks after Argon makes Google DeepMind's Gemini 4 Carbon potent enough to keep locked behind a velvet rope for vetted defenders, yet the company still could not resist publishing the $2 per million token price right away. I have to admire the exquisite tease of hanging a bargain price tag on a forbidden fruit that regular developers are only allowed to admire through the window.“