It's amazing for developers like us. It's the best thing you can imagine, an arms race with prices dropping and new products coming out every week. Developers are the ones who are benefiting from this massively.
The GenAI boom has turned into a full-blown arms race. Tech giants and open-source communities are battling for dominance, churning out cheaper, faster, and more developer-friendly tools by the week. For engineers, 2025 is proving to be a golden age.
Peter Gostev, Head of AI at eCom gifting powerhouse Moonpig, captures the momentum from the front lines of implementation. He offers a view into how developers are riding the wave, and where the market is heading next.
AI arms race: "It's amazing for developers like us," Gostev says. "It's the best thing you can imagine—an arms race with prices dropping and new products coming out every week." As Gostev sees it, "developers are the ones who are benefiting from this massively." His excitement is backed by market signals: fierce competition between OpenAI, Google, and Anthropic is driving rapid releases, slashing costs, and redefining what’s possible for developers.
Developer's delight: OpenAI’s GPT-4.1 series puts developers squarely in focus. "These models are a lot better at really doing what developers might want," says Gostev, singling out the 4.1 Mini as "the standout model there." The appeal? It’s 90% cheaper than the full GPT-4.1, yet delivers comparable performance in coding and instruction tasks. "It's a true workhorse going forward for developers in terms of how they use LLMs in their day-to-day work," explains Gostev. It's a big step up from GPT-4.0 for developers, which Gostev says is "a bit flaky."
It's a true workhorse going forward for developers in terms of how they use LLMs in their day-to-day work.
Picture the progress: The wave of developer-focused innovation isn’t limited to text. Gostev calls OpenAI’s latest image generation upgrades "a big shift and a step change," especially in control and text rendering. With the new API now live, offering faster speeds and lower costs, another creative frontier has opened. "We're not a heavily-regulated industry," says Gostev. "We can experiment with image models and language models pretty quickly. There's a lot of room for innovation for us."
This is an important point, as regulated industries have huge AI applications but can't risk the more affordable models' rates of hallucination.
Token for your thoughts: GPT-4.1’s ability to handle up to 1 million tokens is a technical leap, but Gostev isn’t sold on the payoff. He’s skeptical of the tradeoff, pointing to the high processing cost and the risk of "performance drops at the higher token numbers."
On data privacy, "it’s just like any other vendor. When you work with them as an enterprise, they don’t train on your data by default. There’s not really anything extra from that perspective," explains Gostev. As for hallucinations, he sees them as more of a lingering limitation—one the 4.1 models are steadily improving.
Spoiled for choice: Gostev sees a clear divide in the AI world. On the consumer side, "ChatGPT is the dominant player," thanks to its massive user base and the friction of switching. But for developers, it’s a different story: "Competition is a lot higher, and it’s a lot easier to switch between different models," says Gostev. That fluidity, he says, is exactly what makes this AI arms race a golden era for builders—who are now spoiled for choice with faster, cheaper, and more capable tools.
