Dan Ostermayer's avatar
Dan Ostermayer
ostermayer@primal.net
npub1gc64...uyek
physician creator of deepmarks.org
Dan Ostermayer's avatar
ostermayer yesterday
if you want a super high quality inexpensive cheap inference provider, groq and deepinfra are unmatched. groq is far faster but has less models and is great for batch jobs.
Dan Ostermayer's avatar
ostermayer 3 days ago
ppq is wonderful but deepinfra is even better. groq is cheap and cerebras is super fast. inference providers hosting open source models are the future. just like there are many VPS providers there will be tons of inference providers racing token costs to commodity pricing with minor value adds.
Dan Ostermayer's avatar
ostermayer 1 week ago
https://arxiv.org/pdf/2604.14228 architecture of claude code - A 7-mode permission system acting as a security harness. - A 5-layer context compaction pipeline to store memory. - A subagent delegation mechanism with strict worktree isolation. - Four different extensibility hooks to manage external tools safely.
Dan Ostermayer's avatar
ostermayer 1 week ago
all the fiddling with openclaw seems to be eliminated with hermes. pretty impressive
Dan Ostermayer's avatar
ostermayer 2 weeks ago
AI probably doesn't make big companies more profitable but it lets little companies and individuals compete with big companies...which eventually turns big companies into small companies who then compete with other small companies that compete with individuals
Dan Ostermayer's avatar
ostermayer 3 weeks ago
i've helped a lot of people on GLP-1s transition off by micro dosing and choosing a real food animal based diet. with micro doses the appetite suppression is subtle so you have to create an environment for habit formation. micro doses create less side effects but give some GLP related help. It's like using elastic bands to do your first pull up until you can successfully do it yourself against gravity.
Dan Ostermayer's avatar
ostermayer 3 weeks ago
creating a RAG system is dirt cheap if you use GROQ everyone calling expensive frontier model API's to query their own data is wasting a lot of money that could've been spent building a better RAG system
Dan Ostermayer's avatar
ostermayer 3 weeks ago
@MartyBent have you heard you talk about moving a business onto an open claw based model try your best to put everything on a wiki because the wiki is by definition relational and can be edited by all of your other employees then you can do RAG from the wiki which is essentially what you're doing with openclaw here's an example of what we created for the entire field of emergency medicine with all the knowledge created by people editing a wiki over 15 years
Dan Ostermayer's avatar
ostermayer 3 weeks ago
any study you read that shows the benefits of GLP1 and GIP agonist medications are just showing you the benefits of moving towards a normal body fat percentage
Dan Ostermayer's avatar
ostermayer 0 months ago
if we anchor on ebola being a viral cause we potentially limit our understanding of this deadly disease. imho this is a hypothesis worth testing
Dan Ostermayer's avatar
ostermayer 0 months ago
if you bookmark my posts, consider organizing your bookmarks in my app @Deepmarks.org all your social media nostr bookmarks plus any websites you bookmark are pulled in from your private key and if you buy a lifetime membership we archive everything for posterity on a blossom server
Dan Ostermayer's avatar
ostermayer 0 months ago
the ebola reservoir is often described as a consequence of latent virus in fruit bats crossing over to human infectivity https://academic.oup.com/bioscience/advance-article/doi/10.1093/biosci/biaf050/8116758 Fruit bats live across Africa, Asia, and Oceania, meaning: "The presence of these animals alone does not explain Ebola outbreaks" the one thing that ebola outbreaks all have in common: heavy industry and mineral mining DRC: Major global producer of cobalt, copper, diamonds, gold; outbreaks have occurred in/near mining areas (e.g., gold mining villages). Guinea, Liberia, Sierra Leone: Significant iron ore, gold, diamonds, bauxite; mining was impacted but continued during the 2014–2016 outbreak. Gabon, Republic of the Congo: Oil, manganese, gold, timber-related extraction; historical gold-mining camp outbreaks. Uganda: Gold, copper mining. South Sudan: Gold, copper, iron ore, limestone Ivory Coast: Gold (major producer), diamonds, manganese, bauxite, nickel Every single place that has experienced an “Ebola” outbreak is also an area of heavy mining and exploitation of raw earth materials All of these metal extractive processes share arsenic use and release in common. The symptoms of arsenic poisoning include every sign and symptom present in "ebola" I think it is very likely that if arsenic levels were routinely tested in ebola patients, a strong correlation would begin to emerge. image
Dan Ostermayer's avatar
ostermayer 0 months ago
imagine if you could create particles that serve as the foundational visual evidence of ebola in a sterile starved culture medium that certainly doesn't contain Ebola. without control samples, virologists can be easily trick themselves into thinking a virus is causing a disease rather than a diseased host producing malformed cytoplasmic vesicles that we think are viruses. image