AI Updates You Probably Missed: New Features in Claude Code, Codex, and Gemini
Keeping up with every AI update while running your own work and projects is genuinely hard. So here is a focused summary of the recent updates that will speed up your coding and make your daily use of models like Claude and Gemini easier — so you can start benefiting from them immediately, without losing hours to research.
Across the past few months, Anthropic, OpenAI, and Google have all shipped a wave of features that most casual users have not noticed — because they arrived quietly inside the tools rather than as headline launches. They fall into three buckets.
For developers, both Claude Code and Codex added a Fast Mode — the same model intelligence routed through a faster path, so iteration stops feeling like waiting. Codex went further with Computer Use, Goal Mode, and Appshots. Claude Code added background sessions, live PR status in the terminal, and a clearer usage system.
For everyday users, the Gemini app changed the most. NotebookLM is now built directly into the app as a Notebooks section. A new custom AI avatar lets you create a digital version of yourself that looks and sounds like you. Personal Intelligence connects your Google apps so answers are grounded in your actual digital life. And Gemini now shows a clear usage page, so you can finally see what you are consuming instead of guessing.
This guide walks through all of it — what each feature does, who it is for, and how to start using it today.
Follow for more:
Course Registration: https://halaqa.app/enrollment?course=start-with-ai
Claude Code: Fast Mode and the Quieter Workflow Upgrades
The headline addition to Claude Code is Fast Mode. You toggle it with /fast, and it routes your requests through a faster inference path that delivers roughly 2.5x higher output speed — the same model intelligence, just without the wait that breaks your coding flow. It now runs Opus 4.7 by default. It is a research preview with premium pricing, so the smart move is to turn it on for rapid iteration and debugging, then turn it off again. It is also fully independent from the effort setting, so you can pair Fast Mode with low effort for maximum speed on simple tasks. Beyond Fast Mode, several quieter upgrades matter just as much day to day. Background sessions let you push a long task with Ctrl+B and keep working — and /resume now lists those background sessions alongside interactive ones. The /btw command lets you ask a quick side question without polluting your main task context. The /model command now changes the model for the current session only, with a dedicated key to set a new default. If your branch has an open pull request, the terminal footer now shows its review status with a colored underline that refreshes every 60 seconds — no more switching to the browser. And extra usage was renamed to usage credits for clearer billing language.
Codex: Fast Mode, Computer Use, Goal Mode, and Appshots
Codex received an even larger wave of updates. Its Fast Mode works the same way — a /fast toggle for quick iteration — powered by GPT-5.3-Codex-Spark, an ultra-fast coding model built for day-to-day work. But the bigger story is how far Codex moved beyond just writing code. Computer Use lets Codex operate your computer alongside you: you name a specific app or let it pick one, and it works without bogging down your whole system. Remote computer use keeps it working even after your Mac locks, controllable remotely through Codex Mobile, with safeguards like short-lived authorization. Goal Mode is no longer experimental — you give Codex an objective and it drives toward it for hours or even days, looping through plan, act, test, and review on its own. Appshots, on Mac, is a small but genuinely useful trick: press both Command keys and the frontmost app window is sent to Codex with a screenshot and its text, giving it instant context without you copying, pasting, or describing anything. Codex also gained a memory system that recalls context from past tasks and suggests proactive actions, plus an in-app browser with batch commenting for building web interfaces.
The Gemini App: NotebookLM Inside, Avatars, and Personal Intelligence
The Gemini app changed more than any other product, and most of it landed quietly. The biggest one for serious users: NotebookLM is now built directly into the Gemini app as a Notebooks section. You no longer switch to a separate tool — you organize your chats and research, upload sources, and query them without leaving the app. The custom AI Avatar feature lets you upload a photo or video and create a digital version of yourself that looks and sounds like you, which can then be placed into generated video scenes. Personal Intelligence connects your favorite Google apps so the answers Gemini gives are grounded in your actual digital life rather than generic — it also powers personalized image generation. Just as important for anyone watching their costs: Gemini now shows a dedicated usage page, moving away from confusing daily prompt limits, so you can finally see exactly what you are consuming and judge whether you are on the right plan. On top of all that, the app got a full redesign called Neural Expressive, a native Mac desktop app, free music generation up to three minutes with Lyria 3 Pro, a Thinking level selector, and in-chat file generation across ten formats.
The Pattern Behind All of It — and How to Take Advantage
Step back from the individual features and a clear pattern appears. All three companies are moving in the same direction at once. First, speed: Fast Mode in both Claude Code and Codex exists because waiting breaks focus, and the labs now treat response speed as a feature worth charging for. Second, grounding and personalization: Personal Intelligence in Gemini, memory in Codex, and NotebookLM inside the Gemini app all point the same way — AI that answers from your real context instead of generic knowledge. Third, transparency: Gemini adding a usage page, and Claude renaming extra usage to usage credits, both reflect a shift toward letting you actually see and control what you consume. The practical takeaway is simple. If you write code, turn on Fast Mode on your next session and try Codex Goal Mode on one repeatable task — the time saved is immediate and measurable. If you are an everyday user, open the Gemini app today, find the Notebooks section, and check your usage page. None of these features require a course or a setup process. They are already in the tools you use. The only thing standing between you and the benefit is knowing they exist — which, now, you do.
Prompt
# AI FEATURE UPDATES — QUICK REFERENCE (MAY 2026) # ════════════════════════════════════════ # CLAUDE CODE # ════════════════════════════════════════ # Fast Mode → toggle with /fast — same Opus intelligence, ~2.5x faster output # Now runs Opus 4.7 by default. Research preview, premium pricing. # Independent from the effort setting (combine /fast + low effort). # Background → Ctrl+B to background a task; claude --bg to start one in background # /resume now lists background sessions; /tasks shows all running # /btw → ask a quick side question without polluting the main task context # /model → now changes model for the CURRENT session only # press d in the model picker to set a new default # Usage Credits → extra usage is renamed to usage credits; /usage-credits (old /extra-usage works) # PR Status → if your branch has an open PR, the terminal footer shows review # state with a colored underline, auto-refreshing every 60 seconds # /vim → vim keybindings for editing your prompt input # Plugins → official plugin directory, 55+ curated plus a community marketplace # ════════════════════════════════════════ # CODEX (OpenAI) # ════════════════════════════════════════ # Fast Mode → /fast toggle; GPT-5.3-Codex-Spark is the ultra-fast coding model # (research preview) for quick day-to-day iteration # Computer Use → Codex can operate your computer alongside you — name an app # or let it choose. Runs without bogging down your whole system. # Remote Use → computer use keeps working after your Mac locks, even remotely # via Codex Mobile, with short-lived authorization safeguards # Goal Mode → no longer experimental — give Codex an objective and it drives # toward it for hours or days (plan, act, test, review loop) # Appshots (Mac) → press BOTH Command keys to send the frontmost app window to Codex # with a screenshot and its text — instant context, no copy-paste # Memory → Codex recalls context from past tasks and suggests proactive # actions (e.g. reply to a coworker comment on a Doc you wrote) # In-app browser → build and annotate web UIs; batch comments to request tweaks # ════════════════════════════════════════ # GEMINI APP (Google) # ════════════════════════════════════════ # Notebooks → NotebookLM is now built INTO the Gemini app as a Notebooks section # organize chats and research; upload sources and query them # Usage Page → Gemini now shows a dedicated usage page (moving away from daily # prompt limits) so you can see consumption and pick the right plan # AI Avatar → upload a photo or video and create a custom avatar that looks # and sounds like you — drop it into generated video scenes # Personal Intelligence → connect your Google apps so answers are grounded in your # real digital life; also powers personalized image generation # Neural Expressive → the redesigned Gemini app — pill-shaped prompt box, new # bottom-sheet upload menu, fluid animations and haptics # Mac App → native macOS Gemini desktop app # Lyria 3 Pro → generate music tracks up to 3 minutes long, free # Thinking Level → choose standard or extended reasoning on Fast and Pro models # File Generation → create files in-chat across 10 supported formats # ─── HOW TO START TODAY ─── # Developer? Toggle /fast in Claude Code or Codex on your next debugging session. # Everyday user? Open the Gemini app, find the Notebooks tab, and check the usage page. # Both? Try Codex Goal Mode or Computer Use on one repeatable task and measure the time saved.