Cognition goes under Review; Obsidian plugs SaaS leaks

Welcome to Runtime! Today on Product Saturday: Cognition rolls out a code-review tool for its Devin agent, Obsidian thinks it has a solution for the recent wave of SaaS integration attacks, 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

What are you doing, Devin: While most people using AI software development tools aren't cranking out lines of code like whatever Steve Yegge is doing with Gas Town, they're probably generating more code than their current review processes were designed to handle. If you ask Cognition, code-review systems haven't changed much in over a decade, and this week it introduced a new review tool for its Devin coding agent.

Devin Review "uses state-of-the-art AI + UX to scale human understanding of ever-more-complex code diffs—whether authored by a human, or an agent," it said in a blog post. The tool organizes diffs — side-by-side comparisons of old code and proposed changes — based on how the changes are connected to make the review easier to read, and Cognition said it can also detect bugs.

An agent is on the line: Whenever a major cloud service or SaaS application goes down, dozens if not hundreds of people join an incident call in order to figure out what went wrong and how to fix it. The collective experience of the people in that room can be the difference between solving the issue quickly or watching it linger for hours, and this week Harness introduced a new AI agent designed to help teams capture that experience for future incidents.

The (awkwardly named) Harness Human-Aware Change Agent "listens to the same conversations your engineers are having — in Slack, Teams, Zoom bridges — and transforms the human story of the incident into actionable intelligence that guides automated change investigation," Harness said in a blog post. The company said the tool can pick up on key phrases in those conversations and surface configuration data and recent changes as the incident unfolds.

Chain of fools: Modern enterprise software depends on integrations with other SaaS providers, but the security of those integrations became a big issue last year after the Salesloft attacks. Obsidian Security came out with new tools this week designed to secure "the SaaS supply chain" by giving security teams new ways to detect and respond to attacks.

That includes helping users understand the full scope of the integration points into their networks (no small task for an enterprise) and showing them how one attack will impact their data. Obsidian CEO Hasan Imam told Silicon Angle that more than half of a given company's SaaS integration points are never actually used, and getting rid of those would allow companies to greatly reduce their attack surface.

Ready or not: By now it's clear that poor data quality is a huge obstacle in front of companies that want to develop AI agents, but Factory AI also thinks that software developers who want to jump on the autonomous coding train need to understand whether their code base can actually handle that kind of action. This week it introduced Agent Readiness, which can examine a code base and determine if it is ready for agentic software development.

"Agent Readiness evaluates repositories across eight technical pillars. Each one addresses a specific failure mode we have observed in production deployments," the company said in a blog post. Those pillars include how fast unit tests run and overall code quality, and the tool then scores the user's code base on a scale of one (don't bother with agents) to five (fire all most of your developers).

Party of two: If you needed any further proof that Databricks and Snowflake sit atop the data world, this week Sumo Logic introduced two apps that give mutual customers a clearer picture of how they are moving data in and out of those platforms. “Databricks and Snowflake are core to so many of our customers’ overall corporate data strategies, especially with the increase in AI usage,” Sumo Logic's Keith Kuchler said in a press release.

The Snowflake app focuses on analyzing login data and data pipelines that aren't working as designed, while the Databricks app was designed to audit how employees are using the platform and detecting unauthorized access. The huge increase in data production and consumption thanks to the AI boom proves once again that simply helping companies understand what is happening across their operating environment can be a lucrative business.

Stat of the week

So what are businesses actually getting out of their investments in AI over the last several years? According to a new report from Deloitte, 66% of companies report some degree of productivity gains but "revenue growth remains an aspiration," with just 20% of companies reporting that AI has lifted the top line.

Quote of the week

"The main goal with shutting down the bounty is to remove the incentive for people to submit crap and non-well researched reports to us. AI generated or not." — curl maintainer Daniel Stenberg, announcing the end of the widely used open-source project's bug bounty program after wasting too much time wading through AI-generated slop.

The Runtime roundup

Microsoft resolved a nearly 11-hour long outage affecting several Microsoft 365 services including Outlook early Friday morning, according to CNBC.

Fidelity Investments settled a lawsuit against Broadcom over its right to renew its previous contract for VMware's software, announcing that "Broadcom's services ​to Fidelity will continue uninterrupted," according to Reuters.

Thanks for reading — see you Tuesday!