How a Researcher "Gave" an AI a Case of ADHD and Doubled Its Brainpower
The human brain is a notoriously messy, inefficient, and brilliant piece of biological machinery. We have spent decades trying to replicate its linear logic with computers, but what if the true key to unlocking artificial intelligence lies not in its focus, but in its “distraction?”
This week, solo researcher Udit Akhouri did something that sounds more like a sci-fi premise than a software update: he gave Claude Code—Anthropic’s cutting-edge AI coding agent—a “case of ADHD.” The result, he claims, is an AI that “thinks 2x better.”
The “Divergent” Engine
Akhouri’s tool, aptly named ADHD, is a skill designed to augment the reasoning of AI agents. In a Reddit post that has since gone viral, he explained that the tool was inspired by the “divergent thinking” processes of the ADHD mind.
Instead of pursuing a single, linear path of logic, an AI equipped with ADHD “fans out” into a “tree of thought.” It simultaneously explores multiple parallel ideas, each under a different “cognitive frame.” The system then scores each branch, prunes the dead ends, and dedicates its resources to deepening the most promising survivors.
“It’s good for brainstorming and planning, not coding,” Akhouri clarifies. He positions ADHD as a “reasoning layer” that helps make architectural choices before a single line of code is written.
Novelty vs. repackaging
While the “ADHD” moniker has certainly captured the industry’s attention, some experts question whether the underlying mechanics are truly new. Dr. Sean Robinson, CTO of Empromptu.ai, described it as a “familiar parallel sampling and selection strategy, but packaged in an interesting way.”
Akhouri, however, argues that the real innovation lies in the tool’s transparency and composability. Unlike the “black box” parallel processing of models like GPT Pro, ADHD is an explicit, readable layer that lives inside the user’s Claude environment. Andrew Moore, CEO of Lovelace AI and a former Google VP, agrees that the “genuinely new idea” is its ability to create diversity in a set of parallel thinkers.
The “2x Better” Debate
Akhouri’s claim that his tool makes Claude “think 2x better” is where the real debate begins. He points to a series of internal evaluations where his tool outperformed the baseline across five out of six engineering problems.
However, outside experts are urging caution. The test was performed on a small sample size of only six problems, and the average score was heavily skewed by a massive win in the “trap detection” category. Furthermore, both the testing method and the “judge” were built on Claude’s own tech stack, raising concerns about “same-stack bias.”
“A ‘2x better’ claim needs more than a few open-ended wins,” argues Noe Ramos, VP of AI Operations at Agiloft. “It needs a validated evaluation set, multiple judges, and evidence that the method improves without just rewarding verbosity.”
The BeingsMag Verdict: The Mindset of the Machine
At BeingsMag, we have extensively covered the links between neurodiversity and creativity. The ADHD mindset—with its capacity for non-linear, hyper-associative thinking—has long been a “secret engine” for innovators and artists. Akhouri’s research is a fascinating attempt to replicate that “chaotic brilliance” inside an LLM sandbox.
Whether the “2x” claim holds up under broader scrutiny is yet to be seen. But the very act of “gifting” an AI with the cognitive framework of a neurodivergent mind represents a monumental shift in how we approach artificial intelligence. We are no longer just teaching machines to think logically; we are teaching them to think humanly, in all of our messy, divergent, and occasionally brilliant glory.
