Posts by dotsie@mastodon.social
 (DIR) Post #B4GjzdQcaJpO1KUqMi by dotsie@mastodon.social
       0 likes, 0 repeats
       
       @ariadne @curiousicae this was ‘’mostly’ fixed sometime in mid 2025 for the best frontier models, I had exactly the same issue and bitched incessantly.I think there’s no public paper explaining what was done to address it unfortunately. I’m very interested to read it but everything I find is nonsense about context length and retention which is not the same.There’s a lot of fun stuff at the edges of this using your own model, implementing CoT and distillation / fine tuning is enlightening.
       
 (DIR) Post #B4GkgxpebFpFkmrlIG by dotsie@mastodon.social
       0 likes, 0 repeats
       
       @ariadne @mirth sorry to reply to you in separate threads but I missed these original posts.You are correct about this. I worked in a tertiary way on something with formal verification and some other sorcery involved doing pretty much exactly this pre-public LLM but way after expert systems (which were before my time).There’s fruit there that pays off, they sold the implementation.
       
 (DIR) Post #B4azisbkNZVsbUvoFU by dotsie@mastodon.social
       0 likes, 0 repeats
       
       @Canageek @linear yep, also standards like MISRA etc.This stuff exists it’s just most people have not been exposed to it unless they work on safety critical systems.The idea that the discipline was OK without that level of rigor or formal verification is a disservice to the world as a whole, imo.AI be damned, the industry is very stupid and held together with nearly nothing and a lot of ICs *and* their management are responsible.CS / Software Engineering as a rigorous discipline is rare.
       
 (DIR) Post #B6h91k6OrZORE5sa3M by dotsie@mastodon.social
       0 likes, 0 repeats
       
       @linear run to the hills by Iron Maiden is fast and excellent and not as difficult as it sounds. Very fun to play.