AI Agenda by Stephanie Palazzolo — The Information

Four times/week (subscribers only)

AI Agenda

By Stephanie Palazzolo

Stay ahead of the technological breakthroughs, startups and leaders in AI with a 14-day free trial of the AI Agenda. After your trial, unlock 25% off on a subscription to The Information for full access to the newsletter, plus all of our original reporting on AI.

By providing your email, you agree to The Information’s Privacy Policy

Recent Editions

Getty Images.

The Five Kinds of Model Routers That Cut AI Costs

By Rocket Drew and Laura Bratton · Jul 2, 2026 7:00am PDT · 1 comment

Before we get to today’s agenda, check out Phoebe’s report Wednesday night that Nvidia has launched a new program to financially backstop customers’ purchases of its AI chips. It’s yet another way Nvidia is leveraging its powerful balance sheet to keep the GPU data center party going.On to the column…As more companies reassess rising prices for... Before we get to today’s agenda, check out Phoebe’s report Wednesday night that Nvidia has...

Getty Images

Customer Support Tickets That Code

By Stephanie Palazzolo · Jul 1, 2026 7:01am PDT

Software developers are starting to combine the two most successful uses of large language models: coding and customer support.Specifically, they’re using AI to automate the process of turning customer complaints about software issues or feature requests into new code. Previously, that could be a long and tedious process, which involved sending... Software developers are starting to combine the two most successful uses of large language...

Art by Clark Miller.

OpenAI Discovers New Way to Cut Inference Costs in Half

By Stephanie Palazzolo · Jun 30, 2026 7:06am PDT · 1 comment

We closely track efforts by Anthropic, Google and OpenAI to get access to more server chips to run their models. But we don’t talk enough about the work these companies are doing to get more juice from the servers they already have.In one previously unreported example, OpenAI engineers earlier this month told some colleagues they had figured out... We closely track efforts by Anthropic, Google and OpenAI to get access to more server chips to...

Art via Getty Images

Why Open-Source Models Are Benefitting From The White House Clampdown

By Stephanie Palazzolo · Jun 29, 2026 7:01am PDT

Art via Getty Images

Why Open-Source Models Are Benefitting From The White House Clampdown

By Stephanie Palazzolo · Jun 29, 2026 7:01am PDT

Open-source AI is emerging as a beneficiary of the Trump administration’s recent rapid-fire policy changes towards AI regulation—and of rising costs. Developers and executives have told me that they’ve been burned by building apps on top of closed-source models like Anthropic's Fable 5, only to see the model withdrawn shortly after its... Open-source AI is emerging as a beneficiary of the Trump administration’s recent rapid-fire...

Art by Mike Sullivan

World Models vs VLAs: The Rift Dividing Physical AI

By Rocket Drew · Jun 25, 2026 7:01am PDT

Tech leaders from Elon Musk and Jensen Huang to a slew of startup founders and their venture-capitalist backers say robotics is headed for a “ChatGPT moment” where AI enables a wide range of physical world tasks. But the industry is increasingly split on how to get there.At issue is which kind of AI models are most promising for powering robots.... Tech leaders from Elon Musk and Jensen Huang to a slew of startup founders and their...

Runpod CEO and cofounder Zhen Lu. Photo courtesy of Runpod.

Cloud Startup Runpod Raises $100 Million, Says It Turned Down Buyout Offers

By Stephanie Palazzolo · Jun 24, 2026 7:00am PDT

By some accounts, the compute crunch of 2026 has become even more dire than the chip crunch of 2023, which saw even venture capital firms becoming quasi-cloud providers temporarily. That’s good news for a subclass of firms that popped up to rent out computing power and help developers run open-source models to power applications. Some people... By some accounts, the compute crunch of 2026 has become even more dire than the chip crunch of...

Art by Mike Sullivan

Why Agent Loops Are Hot

By Stephanie Palazzolo · Jun 23, 2026 7:01am PDT

If you’ve been on the AI side of X or Reddit recently, you’ve probably seen the phrase “agent loops” popping up more than usual. That term refers to a new way to run models and AI agents to improve their performance on longer-running or more vaguely-defined tasks, and it’s been gaining traction with developers in recent months.The basic idea... If you’ve been on the AI side of X or Reddit recently, you’ve probably seen the phrase “agent...

Photo by Brent Lewin/ Getty Images

Why Canva Doesn’t See ChatGPT and Claude As a Threat

By Stephanie Palazzolo · Jun 22, 2026 7:01am PDT

Photo by Brent Lewin/ Getty Images

Why Canva Doesn’t See ChatGPT and Claude As a Threat

By Stephanie Palazzolo · Jun 22, 2026 7:01am PDT

Some businesses that use Anthropic and OpenAI models to develop AI apps for coding, design, legal work and other white-collar tasks are getting worried that those two AI model suppliers might eat their lunch.Design app Figma seems attuned to that kind of threat, especially from Anthropic’s recently launched Claude Design app, as we reported. But... Some businesses that use Anthropic and OpenAI models to develop AI apps for coding, design, legal...

Photo via Getty Images

A Competitor to OpenClaw Emerges

By Stephanie Palazzolo · Jun 18, 2026 7:00am PDT

After the success of OpenClaw, an open-source AI agent software that took the tech world by storm earlier this year, it’s no surprise that copycats are popping up. And some of these copycats are catching up to OpenClaw, by some measures.Hermes, an agent tool from Nous Research, recently eclipsed OpenClaw in terms of new GitHub contributors in... After the success of OpenClaw, an open-source AI agent software that took the tech world by storm...

Photo via Getty Images; art by Mike Sullivan

Amazon’s Lower Cost AI Chip Offerings Hold Appeal for Businesses

By Catherine Perloff · Jun 17, 2026 7:00am PDT

For companies debating which AI server chips to use in their data centers, Amazon appears to have one advantage over Nvidia: price. The cost of using Amazon’s Inferentia2 and Trainium’s second generation chips to run existing AI models, known as inference, can be 80% lower than Nvidia’s H100 chips, for comparable workloads, according to ... For companies debating which AI server chips to use in their data centers, Amazon appears to have...

Donald Trump, left, and Dario Amodei. Photos via Getty Images; art by Mike Sullivan

Is Anthropic Losing Goodwill With AI Researchers?; AGI House Alums Raise $25 Million For New Fund

By Stephanie Palazzolo · Jun 16, 2026 7:01am PDT

Remember the last time Anthropic tussled with the Trump administration, all the way back in February, when a coterie of competing AI researchers and CEOs (even Sam Altman!) quickly came to Anthropic’s defense?That doesn’t seem to be happening this time. In the wake of Trump officials’ Friday night move to effectively ban Anthropic’s most... Remember the last time Anthropic tussled with the Trump administration, all the way back in...

Screenshot via YouTube

AI’s Cost Problem Becomes Front and Center, At Least in Asia

By Qianer Liu · Jun 15, 2026 7:01am PDT

Screenshot via YouTube

AI’s Cost Problem Becomes Front and Center, At Least in Asia

By Qianer Liu · Jun 15, 2026 7:01am PDT

Anthropic can’t catch a break! Leaders of the AI firm are meeting with Trump administration officials in D.C. this week to try to resolve a dispute that blew up on Friday night. The administration said that it would need to cut off access to its most advanced Claude Fable 5 and Claude Mythos 5 models to foreign nationals, which led to the... Anthropic can’t catch a break! Leaders of the AI firm are meeting with Trump administration...

Art via Mike Sullivan/Getty Images

Why Microsoft and Other Anthropic Customers Held Off On Using Claude Fable

By Aaron Holmes and Catherine Perloff · Jun 11, 2026 7:25am PDT

Anthropic’s latest AI is so good at coding tasks that many firms are willing to pay top dollar to use it even as Anthropic effectively raises prices. But some customers are holding off over cybersecurity concerns.In releasing its latest model, Claude Fable, Anthropic started a new policy of holding onto any data that customers feed into the... Anthropic’s latest AI is so good at coding tasks that many firms are willing to pay top dollar to...

Dario Amodei. Photo via Getty Images

Anthropic’s New Model Targets Power Users But Cuts Off AI Rivals

By Stephanie Palazzolo · Jun 10, 2026 7:00am PDT

App developers finally got their first real look at Anthropic’s Mythos Tuesday when Anthropic released a somewhat neutered version of the model, Claude Fable 5, following weeks of breathless chatter about it.Developers who tried the new model said the AI largely deserved the hype. They said it thoroughly outperformed earlier Anthropic models in... App developers finally got their first real look at Anthropic’s Mythos Tuesday when Anthropic...

Craig Federighi speaks on a screen at WWDC 2026. Photo by Josh Edelson/Getty Images)

Anthropic’s Mythos Is Coming Today; Apple Pursues Modest Goals with Siri Revamp

By Stephanie Palazzolo and Aaron Tilley · Jun 9, 2026 7:06am PDT

Big news: Anthropic has a new version of a Mythos-class model coming later today called Claude Fable, according to a person with knowledge of the release. It’s a neutered version of Mythos with safeguards to protect against its misuse either in cybersecurity hacks or the construction of bioweapons, the person said.For those who have been living... Big news: Anthropic has a new version of a Mythos-class model coming later today called Claude...

Art by Mike Sullivan

Anthropic Sounds Alarm on Recursive Self-Improvement

By Rocket Drew · Jun 8, 2026 7:00am PDT

Art by Mike Sullivan

Anthropic Sounds Alarm on Recursive Self-Improvement

By Rocket Drew · Jun 8, 2026 7:00am PDT

Last week, Anthropic boasted that Claude is now writing 80% of Anthropic’s code, which the company presented as a harbinger of a critical leap in artificial intelligence called recursive self-improvement. That’s when models are so capable that they can create the next generation of AI without human involvement. Excitement about recursive... Last week, Anthropic boasted that Claude is now writing 80% of Anthropic’s code, which the...

Andy Konwinski, co-founder of Databricks and Perplexity AI. Photo by Marc Fong

Billionaire Databricks and Perplexity Co-Founder Pitches AI Researchers to Not Work for Big Tech

By Laura Bratton · Jun 4, 2026 7:00am PDT · 2 comments

The billionaire co-founder of Databricks and Perplexity AI, Andy Konwinski, is singularly focused on plugging the years-long drain of talent from academia to Big Tech.He wants to encourage academics to focus on publishing more openly available research, a reaction to the move by frontier AI companies to reduce the amount of AI research they... The billionaire co-founder of Databricks and Perplexity AI, Andy Konwinski, is singularly focused...

Art via Shutterstock/Mike Sullivan

The Unintended Consequences of Codex’s Comeback

By Stephanie Palazzolo · Jun 3, 2026 7:01am PDT · 1 comment

OpenAI’s Codex—which was once pitched as a coding agent for developers and is now being framed as a more general-purpose agent for knowledge workers—is having a moment, as OpenAI chief revenue officer Denise Dresser put it on Tuesday. As I wrote about in this story from Tuesday, lots of developers are making the switch from Anthropic’s more... OpenAI’s Codex—which was once pitched as a coding agent for developers and is now being framed as...

Art by Mike Sullivan

Cognition Aims to Be the Switzerland of AI Agents with App Makeover

By Rocket Drew · Jun 2, 2026 7:01am PDT

AI coding startup Cognition is revamping its app to help users manage coding agents from many providers, executives told The Information. Cognition is betting that being a neutral platform will strengthen its hand in the increasingly competitive market for AI coding tools.The change, to be announced on Tuesday, involves the Windsurf application... AI coding startup Cognition is revamping its app to help users manage coding agents from many...

Art via Getty Images

AI Evaluators Struggle with Models That Know When They’re Being Tested

By Rocket Drew · Jun 1, 2026 7:00am PDT · 1 comment

Art via Getty Images

AI Evaluators Struggle with Models That Know When They’re Being Tested

By Rocket Drew · Jun 1, 2026 7:00am PDT · 1 comment

AI researchers are starting to make progress on a confounding problem: AI models are getting better at telling when they are in an evaluation.That could become a problem for AI companies that use evaluations to gauge the capabilities and behaviors of their models before releasing them. If models act differently during testing, that could mean... AI researchers are starting to make progress on a confounding problem: AI models are getting...