You thought training your puppy not to eat your socks was hard.

Michelle Pokrass, post-training lead at OpenAI, has the challenge of fine-tuning the reward mechanisms that steer the company’s large language models (LLMs) to the most desirable answers.

OpenAI’s latest model, GPT-4.1, has features especially geared to help programmers, to ensure they’re getting helpful, usable code, and not half-eaten socks.

The AI lab announced on May 14 that GPT-4.1 would be incorporated into paid tiers of ChatGPT.

“If you want to use 4.1 in ChatGPT, I think it’s a great fit for developers who are asking questions about code or want the model to write some code for them,” Pokrass told us.

But is the code, uh, good?—BH

When PagerDuty CIO Eric Johnson assumed his role at the SaaS digital operations and incident management company two years ago, automation was top of mind.

“There was definitely a need for automation,” Johnson told IT Brew. “I don’t know if the company really knew that. I think they suspected that, but no one was really building a program around that.”

To do so, Johnson and his team built a “small but mighty” crew of staffers with expertise in automation tools to serve as “internal consultants” and determine areas of opportunity for automation at the company, which has more than 1,000 employees across the world. Johnson, who previously served as CIO at SurveyMonkey and Docusign, said he assembled the dedicated team of three employees for the initiative after seeing past automation projects remain stagnant due to a lack of ownership around them.

“Automation was one of those things that was talked about and it was [a] best effort from the existing teams,” Johnson said. “What…I continued to be frustrated by was the fact that we weren’t making real progress on it because it wasn’t someone’s core job.”

See what it did for the help desk.—BM

There was a time at Veterans Affairs when processes called for more paper than code, but times, and applications, have changed.

The Memorial Benefits Management System (MBMS), a web-based app deployed in 2019, for example, supports families and veterans with burial benefits. The task had once required manual data inputs and paper-intensive records, according to Veterans Affairs’ Jacqui Nissen, an acting director for low code/no code in its digital transformation center.

Now the MBMS tracks over 135,000 internments annually, Nissen said, and the VA’s app users don’t have to code to do it. VA agents use modules and menus to pull veteran names from databases and see schedules for the country’s 156 national cemeteries, all integrated within a cloud-based platform-as-a-service (PaaS).

“It has to be fast, seamless, and accurate, so that those family members are not having to wait on a call for an hour and a half,” Nissen said.

Nissen says the VA now has over 200 projects that take a hard PaaS approach and require fewer coding requirements. Plenty of organizations have considered this “low code” method—one with democratizing upside and restrictive drawbacks, according to industry analysts who spoke with us.

Why low-code beats “keyboard pounding.”—BH