The browser becomes the workspace

Tuesday, May 26 · Live from Da Nang

We are running a hands-on Agentic AI workshop in Da Nang.

The workshop is streamed live - join online from anywhere, free of charge.

Sponsored spotlight: Oniichan, a text-to-manga and anime creation tool for character sheets, speech bubbles, and multi-page comic workflows.

TLDR

- OpenAI expanded Codex into Chrome workflows, and now says more than 4 million people use Codex every week.

- Anthropic and Amazon announced a deeper compute partnership, while reports said Google prepared a much larger Anthropic commitment.

- GPT-5.5 and Gemini 3.1 Pro both shipped upgrades that make Agentic AI coding and deep research more practical.

This month's briefing

1. OpenAI brought Codex into Chrome workflows

OpenAI now documents a Codex Chrome extension that lets Codex test web apps, work with browser context, and use DevTools without taking over the whole browser. OpenAI also says more than 4 million people now use Codex every week. The message is clear: AI agents are moving from standalone apps into the workflow.

Sources: OpenAI Developer Docs · OpenAI, Work with Codex from anywhere

2. Reports pointed to a much bigger Google-Anthropic tie-up

On April 24, multiple outlets reported that Google was preparing a commitment of up to $40 billion to Anthropic, including cash and compute. Anthropic also presented Claude on Google Cloud's Vertex AI at Google Cloud Next. Treat the exact deal terms as reported figures, but the direction is clear: cloud capacity has become one of the most important battlegrounds in AI.

Sources: TechCrunch report · Anthropic at Google Cloud Next

3. GPT-5.5 entered the API with stronger Agentic AI capabilities

OpenAI's GPT-5.5, released April 24, is now available in the API. The company says it plans better, uses tools more reliably, and verifies its own output with less hand-holding. Benchmarks show gains on OSWorld-Verified (78.7%) and GDPval (84.9%), both tests of real-world agent performance. For developers, the point is that autonomous coding and research workflows are becoming production-ready.

Source: OpenAI, Introducing GPT-5.5

4. Amazon and Anthropic expanded their compute partnership

On April 20, Anthropic said Amazon was investing $5 billion immediately, with up to $20 billion more possible in the future. Anthropic also committed more than $100 billion over ten years to AWS technologies and secured up to 5GW of capacity for training and deploying Claude. This is less a normal funding round than a long-term infrastructure map.

Source: Anthropic, Amazon compute partnership

5. Enterprise copilots kept moving toward longer workflows

Microsoft's May 2026 Copilot coverage points to GPT-5.2 in Copilot Chat, multi-agent workflows through Copilot Studio, and richer connector understanding for enterprise data. The useful takeaway is broader than one release note: AI products are being sold less as chat windows and more as workflow systems.

Source: Microsoft Copilot Blog, Copilot Studio updates

THE NUMBER

4 million

That is the number of weekly Codex users OpenAI reported in May 2026. The signal is not just adoption, it is repeat usage. Developers are not trying Codex once and leaving. They are building with it.

Source: OpenAI, Work with Codex from anywhere

EDITOR'S PICK

OpenAI Codex for Chrome

If you write code or manage web projects, this is the most practical launch of the month. The extension lets Codex work with browser context, test web apps, and access DevTools without taking over your machine. It is the closest thing yet to an AI agent that fits into an existing workflow instead of replacing it.

ONE THING TO TRY

Run a browser-based coding task with an AI agent this week:

1) Open a web project you are working on, invite an AI agent into the browser, and ask it to fix a real bug or add a small feature.

2) Watch where it succeeds and where it needs guidance. The gap between the two is your personal AI-agent skill map.

May's launches make step one surprisingly fast. The skill map is what turns that speed into lasting productivity.

YOU ASKED, WE ANSWERED

"With all these new models, which one should I actually use?"

The short answer: start with the workflow, not the model. For browser-based coding and web testing, Codex with the new Chrome extension is now the strongest option. For research-heavy tasks that need structured output and source verification, Google Deep Research Max with Gemini 3.1 Pro is hard to beat. For general knowledge work and writing, GPT-5.5 in ChatGPT or the API is the safest default.

Builder tip: the model you choose matters less than how you frame the task. Spend 10 minutes writing a clear prompt template with context, constraints, and desired output format. That single habit will outperform model-hopping every time.

IF YOU READ ONE THING

OpenAI's GPT-5.5 system card

This is the best single technical read from May if you want to understand what "Agentic AI" actually means in practice. The card explains how the model plans, uses tools, and checks its own work. It is not light reading, but it clearly shows why 2026 is the year AI stops being a chatbot and starts being a teammate.

Live Workshop · Tuesday, May 26

Build your first Agentic AI workflow - join us live.

On Tuesday, May 26, we are running a three-hour hands-on workshop on building AI agents and connecting them to real-world tasks. Join from wherever you are - reply to this email and we will send you the access link and slides before we start.

In Da Nang in person? Webe Coffee, 14-16 Nguyen Huu Thong, An Hai, Son Tra.

SPONSORED SPOTLIGHT

Oniichan

Oniichan is a text-to-manga and anime creation tool focused on character design, reusable character sheets, and more consistent multi-page storytelling. It also supports speech bubbles and comic-style page workflows, and it is free to start.

Explore Best-AI.org

Domain switch. Nothing to worry about: We now run on best-ai.org. The old address, best-ai-tools.org, still works and all links lead to the same place. We are not collecting any data twice. Your profile, favorite lists, and saved tools are all preserved.

For security, you may have been logged out. Log in again and you should see your lists as usual.

Built an AI tool?

Submit it to the directory and join thousands of creators on best-ai.org.