Forget vibe coding: Time for vibe docs?

Welcome to Runtime! Today: Factory's co-founder and CEO explains why using AI to automate developers' least-favorite tasks could have a huge impact, AWS sheds more light on the root causes of Monday's outage, and the latest moves in enterprise tech.

(Please forward this email to a friend or colleague! And if it was forwarded to you, sign up here to get Runtime each week.)

Assembly lines

Coding is the most fundamental aspect of a software developer's job, but it's only one of many tasks on their to-do lists. Developers must write testing suites for that code, create documentation so that code can be maintained over time, and figure out how to migrate it to other environments without breaking anything, and that stuff is much less fun.

"Most developers weren't hired because they're really good at writing docs or because they're really good at writing tests, [but] because they have good systems thinking; they can understand the constraints of the problems that they're working on and figure out what are optimal solutions," said Matan Grinberg, co-founder and CEO of Factory, in a recent interview with Runtime. Factory just raised $50 million in Series B funding last month to expand the scope of what its AI-powered "droids" can tackle during the software development process.

Factory's droids are basically AI agents, tools that are capable of autonomously executing tasks in a chain in response to some sort of input, such as creating documentation in response to a pull request.
"The larger an engineering org gets, the less time engineers spend on coding and the more time they spend on things like testing, review, documentation, design docs, meetings, … those are the things that developers don't enjoy doing," Grinberg said.
Another thing developers don't enjoy is getting called into an incident-response meeting when something goes wrong with their code, but Factory's droids can also help coordinate those frantic moments in Slack or Microsoft Teams by surfacing context and zeroing in on potential causes.

It's not that Factory is ignoring agentic coding, which is probably the most promising use-case for large-language models in the enterprise. "Our biggest use case is coding, for sure, but I think one thing to note is that coding from scratch is where all the tools are the least differentiated," Grinberg said.

Factory's droids work with all the popular code editors and IDEs, including Microsoft's Visual Studio Code and Jetbrains, as well as the terminal interfaces that some developers prefer.
However, unlike a lot of new popular coding editors like Cursor and Loveable, Factory started off selling its product to enterprise teams and only just recently released a version of its tools for solo developers.
Grinberg thinks that gives Factory an edge inside companies that already have large code bases and years of documentation stored across various tools: "We create a semantic understanding layer that goes across these different information sources, which is similar to how a human might think," he said.

Engineering managers worried about the quality of AI-generated code might not be terribly enthused about the prospect of automating the process of creating documentation, which plays a far greater role in the long-term success of a software development organization than is generally understood. However, it's not like new hires show up on Day One understanding how to follow a company's best documentation practices either, Grinberg argued.

Those folks are given access to older, vetted documentation and their work is reviewed in regular meetings until leadership decides they can be trusted, and even then that work is still checked over time.
"The more you give in terms of guidance, the better the output will be and the more adherent to whatever you're expecting," he said.

Grinberg believes that successfully automating the boring parts of a software developer's job will make them more likely to actually use AI tools in general, which could deliver the productivity boost that AI coding vendors and buyers desperately hope comes along with investment in these tools.

"You go to an enterprise and you automate all of the coding, there's still going to be so much work that developers do, it's just going to be the least-fun work. So they're not going to want to adopt whatever tool it is anyway," he said.

Dynamo Down Bad

AWS released a detailed summary Wednesday evening chronicling what went wrong inside its notorious us-east-1 region during the early morning hours of Monday, an outage that rippled through the global internet. As stated in its initial communication to customers, the problems started after errors in an automated DNS system used to manage its DynamoDB service created a "race condition," which Akamai defines as a situation in which "multiple processes or threads are simultaneously trying to modify and retrieve shared data, resulting in unforeseen and unintentional outcomes."

In this case, those outcomes led to a situation in which "all systems needing to connect to the DynamoDB service in the N. Virginia (us-east-1) Region via the public endpoint immediately began experiencing DNS failures and failed to connect to DynamoDB," and that included internal AWS systems. Those errors caused a cascading set of issues that ensnared the all-important EC2 compute service, which needed a functioning DynamoDB to work properly, and Network Load Balancer, which needed a functioning EC2 to work properly.

AWS said it disabled the automated DNS system managing DynamoDB and promised to "fix the race condition scenario and add additional protections to prevent the application of incorrect DNS plans" before re-enabling that system. It also said it would "look for additional ways to avoid impact from a similar event in the future, and how to further reduce time to recovery," which is really the key issue here; all cloud infrastructure services will fail from time to time, but this one took way too long to fix.

Enterprise moves

Ragy Thomas is the new co-CEO of UnifyApps, joining Pavitar Singh at the helm of the enterprise AI company.

Karthik Narain is the new chief product and business officer at Google Cloud, joining the company after a decade in technology leadership roles at Accenture.

Brian Emerson is the new chief product officer at New Relic, joining the company after serving as group vice president and general manager at ServiceNow.

The Runtime roundup

Anthropic announced a new computing deal with Google Cloud that "is worth tens of billions of dollars and is expected to bring well over a gigawatt of capacity online in 2026," the company said in a blog post/press release type thing.

IBM's stock recovered from an opening plunge Thursday after it reported revenue and earnings Wednesday that beat expectations but appeared to disappoint investors looking for more growth.

SAP's stock, meanwhile, rose slightly Thursday after missing revenue targets for its third quarter, which further reinforces the idea that stock traders have more in common with (alleged) Portland-area gambling junkies than anything else.

Intel beat expectations for third-quarter revenue despite data-center revenue falling 1%, which it said it hopes to turn around following its deal with Nvidia last month.

Thanks for reading — see you Saturday!