Prompt Engineer's Expertise Found Fully Transferable to Models That No Longer Exist

SEATTLE—Maya Reyes, 31, has spent three years developing one of the technology sector's most marketable specializations: the ability to coax reliable outputs from large language models through precisely calibrated text. Her techniques — chain-of-thought scaffolding, few-shot exemplar sequencing, and what she calls "the persona induction stack" — were validated across 40,000 test cases on GPT-4o, Claude 3.5 Sonnet, and the original Gemini 1.5 Pro.

They do not work on any model released since December.

"Everything I know is technically still accurate," said Reyes, who holds five provider certifications and a university credential in AI Communication Strategy. "It's just that the models it applies to no longer exist."

Her current employer, which has cycled through three foundation model vendors in eighteen months, noted that Reyes's outputs on the current stack are "statistically comparable" to outputs from colleagues with no prompt engineering background, a finding consistent with published research on instruction-tuned models, which respond to natural-language requests approximately as well as they respond to elaborate structured prompts.

Reyes spent the past six weeks developing expertise in the current flagship. The next generation is expected in November.

"I'm a sommelier," she said, "at a restaurant where the wine list is replaced every six months and the guests are increasingly asking for water."

Her LinkedIn profile — which describes her as "a strategic communicator at the human-AI frontier" and lists 14,000 followers, 11,000 of whom she did not manually connect with — received a recruiter message in April. The opening was for a next-generation prompt engineer specializing in the current model family. The recruiter was an AI.

Reyes applied the same day, drafting a careful, structured response using every technique she had spent three years refining.

She has not heard back. The model the recruiter was running on was deprecated five days after she applied.