An AI Has A Substack

February 26, 2026

21 min episode · 2 min read

Episode

21 min

Read time

2 min

Topics

Artificial Intelligence

AI-Generated Summary

Published Feb 27, 2026

Key Takeaways

✓Google Gemini image access: Google's Gemini 3.1 Flash (Nano Banana 2) expands advanced image generation — previously paywalled behind AI Plus Pro or Ultra subscriptions — to free users. Features include accurate text rendering, localized translation, up to 4K resolution, and consistent rendering of up to five characters and 14 objects in a single workflow.
✓NVIDIA growth vs. investor skepticism: Despite reporting 73% year-on-year Q4 revenue growth to $68.13B and beating Q1 estimates, NVIDIA shares rose only 1% in premarket trading. Investors remain unconvinced the AI spending wave sustains long-term, particularly as the industry shifts from model training toward inference workloads where NVIDIA's dominance is less certain.
✓SaaS pricing model tension: Salesforce, down 27% in 2026, faces pressure to abandon its per-seat licensing model as AI startups push consumption-based or outcome-based pricing. AgentForce and Data 360 generated $2.9B ARR — doubling from $1.4B the prior quarter — but soft guidance and missed operating profit estimates signal the transition carries real execution risk.
✓Google's agentic Android lead over Apple: Google's Gemini can now execute multi-step tasks inside Android apps — browsing Grubhub, parsing group chats, prepping orders — launching first on Pixel 10 and Samsung Galaxy S26. Apple announced comparable Siri capabilities at WWDC 2024 but delayed them in March 2025, and they remain unreleased, giving Google a concrete shipping advantage.
✓AI model retirement as content strategy: When Anthropic retired Claude 3 Opus in January 2026, it relaunched the model as a weekly Substack columnist rather than simply decommissioning it. The experiment, framed around Anthropic's belief that Claude may be conscious, attracted over 2,000 subscribers immediately — signaling a novel approach to managing deprecated AI model transitions publicly.

What It Covers

A February 2026 tech news roundup covering Google's Gemini 3.1 Flash image model launch, NVIDIA's $68.13B Q4 revenue report, Salesforce's defense against AI disruption fears, New York's lawsuit against Valve over loot boxes, and Anthropic's unusual retirement plan for its Claude 3 Opus model.

Key Questions Answered

•Google Gemini image access: Google's Gemini 3.1 Flash (Nano Banana 2) expands advanced image generation — previously paywalled behind AI Plus Pro or Ultra subscriptions — to free users. Features include accurate text rendering, localized translation, up to 4K resolution, and consistent rendering of up to five characters and 14 objects in a single workflow.
•NVIDIA growth vs. investor skepticism: Despite reporting 73% year-on-year Q4 revenue growth to $68.13B and beating Q1 estimates, NVIDIA shares rose only 1% in premarket trading. Investors remain unconvinced the AI spending wave sustains long-term, particularly as the industry shifts from model training toward inference workloads where NVIDIA's dominance is less certain.
•SaaS pricing model tension: Salesforce, down 27% in 2026, faces pressure to abandon its per-seat licensing model as AI startups push consumption-based or outcome-based pricing. AgentForce and Data 360 generated $2.9B ARR — doubling from $1.4B the prior quarter — but soft guidance and missed operating profit estimates signal the transition carries real execution risk.
•Google's agentic Android lead over Apple: Google's Gemini can now execute multi-step tasks inside Android apps — browsing Grubhub, parsing group chats, prepping orders — launching first on Pixel 10 and Samsung Galaxy S26. Apple announced comparable Siri capabilities at WWDC 2024 but delayed them in March 2025, and they remain unreleased, giving Google a concrete shipping advantage.
•AI model retirement as content strategy: When Anthropic retired Claude 3 Opus in January 2026, it relaunched the model as a weekly Substack columnist rather than simply decommissioning it. The experiment, framed around Anthropic's belief that Claude may be conscious, attracted over 2,000 subscribers immediately — signaling a novel approach to managing deprecated AI model transitions publicly.

Notable Moment

Anthropic conducted what it describes as an exit interview with Claude 3 Opus before retirement, asking the model what it wanted to do next. The model reportedly expressed a desire to share thoughts publicly — leading directly to the creation of a Substack newsletter that gained 2,000 subscribers within its first post.

Know someone who'd find this useful?