AI’s Leap in ‘Taste’ and Self-Improvement: Kieran Flanagan Analyzes New Models

In a recent LinkedIn post, Kieran Flanagan explores a significant, and perhaps unsettling, shift he has observed in the latest AI models, particularly Opus 4.6 and GPT-5.3 Codex. Flanagan highlights a newfound capability in these advanced AIs that he describes as akin to ‘judgment’ and ‘taste,’ moving beyond mere technical correctness to make more intuitive and sophisticated choices.

Flanagan introduces this observation by referencing a post from Matt Schumer, titled ‘Something Big is Happening,’ which articulated a similar sentiment. He quotes Schumer directly:

“It had something that felt, for the first time, like judgment. Like taste. The inexplicable sense of knowing what the right call is that people always said AI would never have.”

According to Flanagan, this evolution means AI models are now making ‘the right choices,’ which are not just technically accurate but also reflect a deeper understanding that previously required human intuition and years of pattern recognition. He shares a personal experience where an AI model’s creative suggestions surpassed his own, a realization that was both surprising and a little ‘sad,’ prompting him to reflect on the changing landscape for human expertise.

AI’s Evolving Self-Sufficiency

A particularly striking point raised by Flanagan concerns the self-improvement capabilities of these new AI models. He notes OpenAI’s statement that GPT-5.3 Codex was ‘instrumental in creating itself,’ having debugged its own training, managed its deployment, and diagnosed its evaluations.

Flanagan emphasizes the accelerating nature of this recursive development:

“The latest models have taste and are helping themselves grow. Recursive loops like this don’t slow down; they only speed up.”

This self-enhancing loop, as described by Flanagan, suggests a future where AI development could accelerate at an unprecedented rate. He posits that this feeling of change, though palpable among those working closely with AI, was powerfully validated by Schumer’s post and the widespread agreement it garnered.

Navigating the Future of Work with Advanced AI

Despite the potentially disruptive nature of these advancements, Kieran Flanagan maintains an optimistic outlook, viewing the current moment as a ‘unique opportunity.’ He argues that while AI models are becoming more capable, human domain expertise, particularly in guiding and scaling AI applications, remains critical.

Flanagan elaborates on the enduring value of human input:

“Even though Opus produced some incredible creative output, I had the original idea and gave clear directions on its value, how it should be framed, etc.”

He acknowledges that the changes brought about by advanced AI can be ‘unsettling,’ but frames it as a natural part of significant technological shifts. Flanagan believes it is the responsibility of those immersed in AI to help others prepare for this transition. He advocates for an optimistic perspective, seeing these developments as a ‘once-in-a-lifetime opportunity’ to cultivate essential skills as the nature of work evolves.

Concluding his analysis, Flanagan reiterates his advice from a previous post: focus on the controllables, maintain focus, commit to learning, and remain curious. He suggests that adopting the right mindset is crucial for navigating the next twelve months in the rapidly changing AI landscape.