Google apps in the terminal

2026-03-05 · Source: Ben's Bites · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, extended

Summary

Google has released Gemini 3.1 Flash Lite, a fast model outperforming Haiku 4.5 on benchmarks, though at a higher price point of $0.25/$1.50, making open-source alternatives like Minimax M2.5 more competitive for developers seeking stronger performance over speed. OpenAI updated ChatGPT's default model to GPT-5.3-Instant, featuring improved behavior with reduced hallucinations and better web search. Additionally, OpenAI's Codex is now available on Windows, and the company is reportedly developing an internal GitHub alternative, exploring ads in ChatGPT with The Trade Desk, and preparing for an IPO, with its annualized revenue reaching $25 billion, slightly ahead of Anthropic's $19 billion. Google Workspace also introduced a CLI with a focus on agents for services like Drive and Gmail, alongside a blog post on rewriting CLIs for AI agents. The content also highlights various AI tools, applications, and personal AI infrastructure setups, including local model hosting and agentic workflows.

Key takeaway

For AI Engineers evaluating model deployment strategies, consider the trade-off between Google's Gemini 3.1 Flash Lite's speed and its increased cost versus the performance of open-source models like Minimax M2.5. Your choice should align with specific project requirements, prioritizing either rapid inference or raw computational power. Additionally, explore the new Google Workspace CLI and agent-focused design principles to enhance automation and integration within your development workflows.

Key insights

New models and tools enhance AI agent capabilities, driving both commercial and open-source innovation.

Principles

• Prioritize performance or cost based on application needs.
• Agent-focused design improves CLI and application utility.

Method

A personal AI infrastructure can integrate local models, cloud subscriptions, and agent harnesses for diverse tasks like coding, data analysis, and browser automation.

In practice

• Use VLM for faster local model inference.
• Employ Vibe CLI to unify AI model access.
• Leverage Super Whisper for voice-to-text AI interaction.

Topics

• AI Agents
• Local LLM Infrastructure
• Generative AI Models
• AI Development Tools
• AI Business & Trends

Code references

• googleworkspace/cli
• ygwyg/agent-experience
• paperclipai/paperclip
• nolenroyalty/yt-browse
• openai/symphony

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Google's Managed Agents API promises one-call deployment at the cost of execution layer control — Google's Managed Agents centralize AI agent orchestration and execution, simplifying deployment but raising concerns about control and predictability.
• LWiAI Podcast #236 - GPT 5.4, Gemini 3.1 Flash Lite, Supply Chain Risk — Major AI players are rapidly advancing model capabilities and agent integration while navigating significant ethical and business challenges.
• What Google Needs to Do at I/O This Week — AI is diverging into "abnormal" work-focused agent management and "normal" consumer-facing applications, shifting human roles from execution to supervision.
• Why Google Workspace CLI is a Big Deal — CLIs are becoming the preferred interface for AI agents due to their low friction and deterministic output.
• 😸 OpenAI, Gemini, Qwen new models — The AI industry is shifting towards smaller, faster, and more cost-effective models for broad deployment.
• EP219: 12 Open-source LLMs — Open-source LLMs offer diverse capabilities, while model and agent architecture choices depend on task complexity and resource constraints.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ben's Bites.