Developers are still weighing the pros and cons of AI coding agents

Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week via email here.

Is ‘AI slop’ code here to stay?

A few months ago I wrote about the dark side of vibe coding tools: they often generate code that introduces bugs or security vulnerabilities that surface later. They can solve an immediate problem while making a codebase harder to maintain over time. It’s true that more developers are using AI coding assistants, and using them more frequently and for more tasks. But many seem to be weighing the time saved today against the cleanup they may face tomorrow.

When human engineers build projects with lots of moving parts and dependencies, they have to hold a vast amount of information in their heads and then find the simplest, most elegant way to execute their plan. AI models face a similar challenge. Developers have told me candidly that AI coding tools, including Claude Code and Codex, still struggle when they need to account for large amounts of context in complex projects. The models can lose track of key details, misinterpret the meaning or implications of project data, or make planning mistakes that lead to inconsistencies in the code—all things that an experienced software engineer would catch.  

The most advanced AI coding tools are only now beginning to add testing and validation features that can proactively surface problematic code. When I asked OpenAI CEO Sam Altman during a recent press call whether Codex is improving at testing and validating generated code, he became visibly excited. Altman said OpenAI likes the idea of deploying agents to work behind developers, validating code and sniffing out potential problems. 

Indeed, Codex can run tests on code it generates or modifies, executing test suites in a sandboxed environment and iterating until the code passes or meets acceptance criteria defined by the developer. Claude Code, meanwhile, has its own set of validation and security features. Anthropic has built testing and validation routines into its Claude Code product, too. Some developers say Claude is stronger at higher-level planning and understanding intent, while Codex is better at following specific instructions and matching an existing codebase.

The real question may be what developers should expect from these AI coding tools. Should they be held to the standard of a junior engineer whose work may contain errors and requires careful review? Or should the bar be higher? Perhaps the goal should be not only to avoid generating “slop” code but also to act as a kind of internal auditor, catching and fixing bad code written by humans.

Altman likes that idea. But judging by comments from another OpenAI executive, Greg Brockman, it’s not clear the company believes that standard is fully attainable. Brockman, OpenAI’s president, suggests in a recently posted set of AI coding guidelines that AI “slop” code isn’t something to eliminate so much as a reality to manage. “Managing AI generated code at scale is an emerging problem, and will require new processes and conventions to keep code quality high,” Brockman wrote on X.

Saas stocks still smarting from last week’s ‘SaaSpocalypse’

Last week, shares of several major software companies tumbled amid growing anxiety about AI. The share prices of ServiceNow, Oracle, Salesforce, AppLovin, Workday, Intuit, CrowdStrike, Factset Research, and Thompson Reuters fell so sharply that Wall Street types began to refer to the event as the “SaaSpocalypse.” The stocks fell sharply on two pieces of news. First, late in the day on Friday, January 30, Anthropic announced a slate of new AI plugins for its Cowork AI tool aimed at information workers, including capabilities for legal, product management, marketing, and other functions. Then, on February 4, the company unveiled its most powerful model yet, Claude Opus 4.6, which now powers the Claude chatbot, Claude Code, and Cowork.

For investors, Anthropic’s releases raised a scary question: How will old-school SaaS companies survive when their products are already being challenged by AI-native tools? 

Although software shares rebounded somewhat later in the week, as analysts circulated reassurances that many of these companies are integrating new AI capabilities into their products, the unease lingers. In fact, many of the stocks mentioned above have yet to recover to their late-January levels. (Some SaaS players, like ServiceNow, are now even using Anthropic’s models to power their AI features.)

But it’s a sign of the times, and investors will continue to watch carefully for signs that enterprises are moving on from traditional SaaS solutions to newer AI apps or autonomous agents.

China is flexing its video models

This week, some new entrants in the race for best model are very hard to miss. X is awash with posts showcasing video generated by new Chinese video generation models—Seedance 2.0 from ByteDance and Kling 3.0 from Kuaishou. The video is impressive. Many of the clips are difficult to distinguish from traditionally shot footage, and both tools make it easier to edit and steer the look and feel of a scene. AI-generated video is getting scary-good, its main limitation being that the generated videos are still pretty short.

Sample videos from Kling 3.0, which range from 3 seconds to 15 seconds, feature smooth scene transitions and a variety of camera angles. The characters and objects look consistent from scene to scene, a quality that video models have struggled with. The improvements are owed in part to the model’s ability to glean the creator’s intent from the prompts, which can include reference images and videos. Kling also includes native audio generation, meaning it can generate speech, sound effects, ambient audio, lip-sync, and multi-character dialogue in a number of languages, dialects, and accents.

ByteDance’s Seedance 2.0, like Kling 3.0, generates video with multiple scenes and multiple camera angles, even from a single prompt. One video featured a shot from within a Learjet in flight to a shot from outside the aircraft. The video motion looks smooth and realistic, with good character consistency across frames and scenes, so that it can handle complex high-motion scenes like fights, dances, and action sequences. Seedance can be prompted with text, images, reference videos, and audio. And like Kling, Seedance can generate synchronized audio including voices, sound effects, and lip-sync in multiple languages. 

More AI coverage from Fast Company: 

• We’re entering the era of ‘AI unless proven otherwise’
• A Palantir cofounder is backing a group attacking Alex Bores over his work with . . . Palantir
• Why a Korean film exec is betting big on AI
• Mozilla’s new AI strategy marks a return to its ‘rebel alliance’ roots

Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.

View the full article