I feel like there is great content in here but the marketing stuff is confusing it. Is this actually some researchers work? Did you cite it anywhere? I tried to read the article and with the context of your message it was confusing. If it's your findings then please tell us, if it's from something else tell us that too. Either way it's value but you started with that it's researchers work and someone released it and then wrote as if it's just your article
This to me explains why Anthropic is saying "remove all your instructions" to work with Opus 5 and others are saying it doesnt play well with workflows. They basically tried to bake the workflow into the model, now its thinking "what do you really mean" and "what is the real or larger problem".
This is what Wispr gets for not innovating. Command mode was such a key feature that stayed weak way too long.
Depends on if its costly to maintain/support or not. If it's not, then your just gambling on when that FCF goes to zero.
Are you using customized handoff and catchup commands? My system is self improving by keeping a log of decisions and learnings in handoff. Then I have a /review command that looks at all the learnings and finds patterns to elevate to rules or systems.
show some initiative, I believe in you. Acct disbursement is a big deal. Presidents give speeches but governments follow budgets. You need a treasury instance that doles out tokens / accts to towns based on funding requests that are granted if the town "did what human wanted"