Noah Sussman

@noahsussman.bsky.social

“Time is an illusion that helps things make sense.” — R. Sugar

A computer cannot tell what is good and what is bad. It can write working code but it can’t know whether it wrote good code or bad code. That’s why AI can never fully replace programmers. In general it’s the reason you can never fully remove the “human in the loop.”

A Metcalfe wheel is the complete connectivity within which a hypercube projection can be inscribed.“​​​​​​​​​​ if you can visualize your communication system as a Metcalf wheel, then there is another way to visualize it: as a multidimensional structure that information traverses

A computer cannot tell what is good and what is bad. It can write working code but it can’t know whether it wrote good code or bad code. That’s why AI can never fully replace programmers. In general it’s the reason you can never fully remove the “human in the loop.”

Why do you think LLMs have a problem solving flaky tests in real-world work situations? Google Scholar and Arxiv are bristling with papers on how to succeed fixing flaky tests with LLMs. So what do you think is the delta? Why isn’t the academic success crossing over into the enterprise?



Have you used an LLM to fix a flaky test?

Yes and it worked!  27%

Yes and it didn't work.  27%

Scientists Finally Reveal Biological Basis of Long COVID Brain Fog. Patients with brain fog may have disrupted AMPA receptor (AMPAR) expression—key molecules for memory and learning. Imaging of Long COVID patients showed elevated AMPAR density tied to more severe cognitive impairment.

Scientists Finally Reveal Biological Basis of Long COVID Brain Fog

Researchers employed a specialized brain imaging technique to identify a potential biomarker and therapeutic target for Long COVID. More than four years after the onset of the COVID-19 pandemic, scien...

scitechdaily.com

whether deliberate or unintentional, the people who want you to treat AI systems as human-like fully autonomous "agents" are simply laundering responsibility and accountability away from those who choose to monetize these systems onto the artifacts that can’t know responsibility or accountability