Salesforce drops the UI; DuckDB hits the lake

Welcome to Runtime! Today on Product Saturday: Salesforce might have the answer to SaaS sustainability in an AI world, DuckDB addresses the "small changes" problem in lakehouses, and the quote of the week.

Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week, and if you value independent enterprise tech journalism, click the button below and become a Runtime supporter today.

Ship it

Headless UI for topless SaaS: The collective stock-market freakout over the future of enterprise software seems to have eased a bit, but lots of questions still remain about how companies will want to interact with so-called systems of record. Salesforce introduced a new "headless" design across several key services this week that could provide a road map for other SaaS stalwarts struggling to figure out their next move.

Salesforce Headless 360 presents "the capabilities your agents need most, exposed as an API, MCP tool, or CLI command so humans and agents can build, act, and deliver experiences on any surface," the company said in a blog post. It's been clear for some time that generative AI is going to change the way people interact with software, and the data — not the user interface — is really the most important part of a service like Salesforce.

Racked and stacked: As code review becomes the bottleneck with the rise of AI coding agents, software teams are going to need help sorting through all that code to make sure it is ready for production. This week GitHub introduced Stacked PRs, which promises to help those teams break down those large review jobs into smaller pieces.

Stacked PRs allows developers to "arrange pull requests in an ordered stack and merge them all in one click," the company said in a blog post. Developers used to have to hack together a workflow for doing this, but it's now supported natively, and "its biggest impact is eliminating rebase hell — the manual effort of updating multiple dependent branches when the base changes," Pareekh Consulting's Pareekh Jain told InfoWorld.

Ducks on the lake: Data lakehouses offer a fast and convenient way to store data in open formats where it can be accessed by different data query engines, but they've suffered through what DuckDB calls the "small files" issue: "The root of the problem is that every small write creates a new data file and updates the metadata," the company said in a blog post this week. That process creates a lot of cruft for query engines to wade through, and DuckDB's solution, called DuckLake, became generally available this week.

DuckLake uses a technique the company calls "data inlining" to store those changes in the data catalog, rather than in storage. "The key insight here is that database systems like PostgreSQL, but also DuckDB and others, are much, much better at handling small changes than object stores," DuckDB CEO Hannes Mühleisen told The Register.

Move fast and protect things: CIOs and CISOs are right to worry about what happens when AI agents are allowed to access sensitive corporate data and take action based on that data, given the potential for that data to spiral out of their control. Commvault introduced three new services this week that promise to let those executives sleep easier by giving them a way to make sure that data is protected from rogue agents.

"Data Activate, AI Protect, and AI Studio will address these challenges by helping teams understand the impact of agent-driven changes and roll back when necessary," the company said in a press release. According to Blocks and Files, the three new services "recognize that large language models and agents need access to safe, reliable data, and that AI agents, like people operating within an IT estate, need their privileges and actions overseen and rolled back if something goes wrong."

Leap of faith: Software teams that have embraced AI coding agents need a deterministic way to make sure those non-deterministic tools are actually shipping code that works. Leapwork introduced a new service called Continuous Validation Platform this week that helps those teams sign off on that code long before it hits production servers and breaks systems in ways that can be hard to diagnose.

"In live customer environments, the platform delivered up to 75% faster test automation implementation, a 50-70% reduction in test maintenance effort and up to a 90% reduction in functional defects reaching production," the company said in a press release. Coding agents are going to change just about every part of the software development life cycle, and new tools are going to be needed at several stages throughout that cycle to manage this new reality.

Stat of the week

Enterprise AI simply doesn't work without organized data, and the cost of making sure data infrastructure is ready for AI tools is going up. According to new research from dbt Labs, 57% of data professionals said they are spending more on storage and compute, but only 36% reported receiving an increase in their budgets to handle that growth.

Quote of the week

"Our Microsoft partnership has been foundational to our success. But it has also limited our ability to meet enterprises where they are – for many that’s Bedrock." — OpenAI chief revenue officer Denise Dresser, explaining to employees just how much the enterprise AI world has turned now that AWS is working with OpenAI.

The Runtime roundup

AI chip maker Cerebras filed to go public Friday, revealing that revenue grew 76% in 2025 to $510 million compared to the previous year, according to CNBC.

Hyperscalers and AI model companies are falling behind on data-center construction amid power shortages and local opposition, according to The Financial Times.

Thanks for reading — see you Tuesday!