Larry Dignan

OpenAI said it is updating ChatGPT's shopping engine with better product discovery. In a blog post, OpenAI said:

"Instead of scrolling through pages of results, ChatGPT can help you find what you’re looking for based on your budget, preferences, and constraints—and surface products that fit. You can browse visually, upload images as inspiration for similar items, and refine results conversationally until you land on the right option."

The upshot here is that OpenAI is going to have to step up its shopping experience if it's going to compete with Google and Amazon at some point for product discovery.

Miro said it will acquire Reforge, which is an AI platform for product teams, in that aims to bridge product development and frameworks with AI-driven collaborative workflows.

Andrey Khusid, CEO of Miro, said in a statement that the deal is aimed at helping customers innovate faster by focusing on what to build as well as the how.

Reforge Learning will continue operate separately at Reforge.com and the combined company will continue to invest in new courses. Brian Balfour, Reforge CEO, will become Miro's Chief Growth Officer. Tom Willerer, Reforge's current COO, will join Miro as Chief Strategy Officer.

• Miro aims to smooth out AI, human collaboration
• Constellation ShortList™ Digital Canvas Workboards
• Constellation ShortList™ Collaborative Product Development Solutions

Claude Cowork and Claude Code now have computer use to complete task on the Mac OS. The feature is available in research preview for Claude Pro and Max subscribers.

Apple said its annual Worldwide Developers Conference (WWDC) will be online from June 8-12. Apple will hold an in-person event during the keynote and state of the platform address June 8.

Time flies in the cloud as AWS just turned 20 years old. The company posted a blog post on the milestone and a bit of history about S3, which started the AWS push. Geekwire posted a good retrospective too.

For giggles, I'll celebrate the AWS 20th by quoting me far too long ago way back in 2008. I said:

Amazon announced persistent storage for its EC2 service and what's notable is how quickly the e-tailer is running ahead of the competition. In fact, Amazon's real business down the line will be its cloud services. Amazon will be like a book store that sells cocaine out the back door. Books will be just a front to sell storage and cloud computing.

Elon Musk announced Terafab, a project run by Tesla and SpaceX to manufacture chips for robotics, AI and space data center processes. Tesla, xAI and SpaceX would all use these processors. This manufacturing facility would be the largest ever at 1TW/year and combine logic, memory and advanced packaging in one place.

Tesla's Optimus robots would help build Terafab.

It's not hard to find skeptics since some of the math behind Terafab stretches the imagination (and possibly reality), but if Musk is going to make space data centers a reality Terafab is a good first step.

Thoughtworks and Teneo formed a joint venture to meld business and AI transformation end to end. The joint venture will focus on AI transformation holistically from boardroom strategy to corporate advisory as well as AI and software engineering services.

The joint venture combines Teneo's more than 1,800 C-suite and board advisors with Thoughtworks' more than 10,000 engineers for AI delivery.

The New York Times did an article on tokenmaxxing, which is the idea that your performance should be measured on the tokens consumed when AI coding. In some cases, the number of tokens used is turning up in performance reviews. Puhleeze. For those of you with a bit of gray hair you'll recall that employees at a few tech giants were graded on how they used internal social networks. Under that logic, your performance was somehow equated to how much you performatively collaborated with others at your company.

Like that enterprise social experiment, which if I recall was all based on Salesforce's Chatter platform, tokenmaxxing will die a swift death as soon as the bills come in. Tokenmaxxing is the equivalent of getting a pat on the back when you left your cloud instance running over the weekend and blew through the budget.

In a long non-apology about Windows 11, Microsoft said it will "be more intentional about how and where Copilot integrates across Windows, focusing on experiences that are genuinely useful and well‑crafted." Windows Insiders have been screaming about how Copilot has landed in cut and paste and ridiculous places like Snipping Tool and Notepad.

Microsoft also said you'll now be able to move the taskbar. And the company pledged to make Windows Updates more predictable and less annoying. To crib an old motto from Ford, Microsoft is making quality job one. Hopefully, Microsoft's quality push will turn out better than Ford, which would run a quality commercial and then recall your car.

WordPress said AI agents can now draft and publish blog posts, build and update pages, manage comments, organize content and update media metadata.

Mehdi Paryavi, CEO of the International data Center Authority, said the following at Constellation Research's AI Forum in Silicon Valley. IDCA is a global digital economy think tank based in Maryland.

• "We need to find partnerships and alliances. We cannot be creating wars for enemies and then expect our economy to grow. The digital economy requires peace. People put their data where there is peace and peace dividend as a standard, very important. It's very important to create bonds."
• "AI is a good instrument for scientific advancement and growth, but it is not something your kids should be playing with day in, day out. There has to be some moral guardrails. There has to be some national guardrails. There has to be some frameworks of thought put around all of this. Otherwise, we're just headed towards a slippery road and the fact that nobody has wrapped their head around what it really means to be AI driven. I mean, people can't write anymore. People can't analyze anymore."
• "We can really optimize. I can take you to 35,000 federal data centers of this country that are obsolete. They're burning energy like nothing. Nobody knows where they are, who's the owner, who's the stakeholder. Unless you unplug them, nobody would know who's the owner. We have issues here, and nobody planning long term."

Cursor said Composer 2 is generally available and is "frontier-level at coding" and outperforms models from OpenAI and Anthropic.

Key points:

• Composer 2 is 10x cheaper than Opus 4.6 and 5x cheaper than GPT-5.4.
• Cursor decided to build its own coding models and built Composer 2 in 9 months.
• Cursor will keep model neutrality and 80% of users leverage at least two model families. Composer is No. 2 most popular model on Cursor and it's likely Composer 2's price and performance will be at least equally as popular.