Every now and then, I think it's useful to look at what great work looked like in its first form.
In retrospect, this result was the first step toward the LLMs that are now the dominant approach in AI.
In 2017, this was a very modest result and probably didn't seem very promising. It took real courage and tenacity on OpenAI's part to keep at it for the next 5 years.
This is the first new paper from Alec Radford since leaving OpenAI. Token-level data filtering is kind of a simple idea, but so are many effective ideas in LLMs.
One advantage is that this type of safety guardrail can't be undone by an adversary in post-training, so it's a good fit for open source models.
The experiments are all done in preventing models from acquiring medical
capabilities, while preserving related capabilities like e.g., biology.
> My doctor said she’s seeing a huge number of patients coming back with excessively high Vitamin D levels after taking supplement doses recommended by influencers. It happened to me, too, with what I though was a conservative dose of Vitamin D (5K IU, not even taken every day)
IMO that’s part of what’s interesting about this study design — they tested vitamin D blood levels and adjusted the supplement dose based on that. This seems like a much better approach than taking a high dose blindly.
I think the headline is accurate. The 52% number is from the experiment arm (participants who received a vitamin D supplement, with the quantity guided by blood testing). While it’s technically possible for the supplement dosage to be calculated as zero, 85% of participants were deficient at baseline, so this isn’t the main effect.
> I think the headline is accurate. The 52% number is from the experiment arm (participants who received a vitamin D supplement, with the quantity guided by blood testing). While it’s technically possible for the supplement dosage to be calculated as zero, 85% of participants were deficient at baseline, so this isn’t the main effect.
Yes, but it's also important to note that the study wasn't on a representative sample of the general population. They recruited people who had acute coronary syndrome. The average age was over 60 years old, 80% were men, and half of them had already had at least one heart attack.
Can you paste an example title and email here? I'm a founder and get 100's of emails per day. I often reply to cold emails, but honestly, with 90% of them, I can't really tell if they will solve one of my top 3 problems.
Totally. Here is what my emails typically look like:
Re: quick scheduling question
Hi [name],
I’ve followed [company] for a while and the scale you’re operating at is impressive. With customer and partner meetings, plus internal meetings, I imagine your calendar gets pretty full.
I’m working on a way for founders to cut down the email back-and-forth of scheduling without hiring or managing an EA.
Is scheduling something you still handle personally at all?
I think your second sentence is the strongest -- it's specific about what you actually help with.
Right now, the title and preview text on mobile probably look like "Quick scheduling question" / "Hi [name], I've followed [company] for...". This looks like a lot of other emails in my inbox that are requesting a meeting time. So I think a lot of people who quickly scan their inbox would pass this by and not even open the full email.
I'd also drop the "re: " part unless it's genuinely a reply. It breaks trust.
Got it. That's fair, I could definitely remove the first sentence. However, I still want the email to seem personal. Do you think it is important to have that personalized part somewhere in the email?
There's no re, I just put that there instead of saying Subject:. Sorry, that's confusing.
I feel this is the wrong business direction. If you do it well, at most you'll only save on the salary of an administrative assistant, but they actually want much more, and this small improvement won't attract their attention.
I get the concern if the value is framed as “replacing an admin assistant”, but for many founders, the real win isn’t salary savings, it’s eliminating context switching, follow-ups, and mental overhead around coordination. As well as eliminating the mistakes that can be made by humans. Even small logistics can disproportionately drain attention. If done well, this is less about cost reduction and more about reclaiming focus and execution speed.
This was a write-up of a new study (TARGET-D) that used vitamin D supplements -- with the supplement amount guided by blood testing -- to reduce heart attack risk.
I've been working in heart health in 10 years and I was surprised at the magnitude of the effect here.
I hope it holds up as they move toward the final publication. Vitamin D supplementation is cheap and this could have a huge benefit.
I’ve had 5 different primary care doctors across multiple practices in different locations due to moving, changing jobs, and doctors retiring
Every single one of them included Vitamin D testing in the annual checkup.
Two of my jobs in the past few years have had wellness programs that offered free Vitamin D testing along with a couple other things (A1c, lipids)
It’s very common in the United States at least. I know this goes against the “US healthcare bad” narrative but one of the difficulties with our costs is that we get more testing and procedures. Cutting those costs is going to be hard because people like the freedom to have their doctor order common tests
The 25-hydroxy vitamin D test (aka: 25(OH)D test) is not part of a lipid panel, comprehensive (nor standard) metabolic panel, or any number of tests I have regularly. Without a specific request, it's unlikely anyone gets tested for this unless maybe you're a psychiatric patient. When I had severe depression in my 20s, a doctor did have this test done.
It's surprising because so many people are deficient and the treatment is extremely cheap. It's bizarre, it's like if a problem is a big enough problem, medicine says well most people are living with it and washes its hands.
People in particular groups with higher risk of deficiency will be tested every year by many doctors. That practice obviously can't amount to testing every year being the average though.
No, not typically. Myself I would usually order one either on specific request, or to investigate things like osteoporosis or pathologic fractures, but not as screening. USPSTF does not currently recommend vitamin D screening either in asymptomatic, non-pregnant adults ( https://www.uspreventiveservicestaskforce.org/uspstf/recomme... ).
Specifically: “ Although our pharmaceutical armamentarium is very good at the moment (the combination of statin-ezetimibe-proprotein convertase subtilisin/kexin type 9 [PCSK9] can reduce LDL cholesterol [LDL-C] levels by 85%), new drugs are emerging through the different pitfalls of current drugs.”
The lack of self-consistency does seem like a sign of a deeper issue with reliability. In most fields of machine learning robustness to noise is something you need to "bake in" (often through data augmentation using knowledge of the domain) rather than get for free in training.
reply