rogerkirkness

Collaborative email, meeting recording, knowledge search and goal tracking in one thing. The search applies across emails and meetings, but also other things. We had to figure out whether it would be sufficiently good to connect third party tools, and basically concluded no. We did some research to understand why (1).

[1]: https://arxiv.org/abs/2504.07106

Keywords or vector search on their own don't get good results for high entropy queries. MCP type approach is good for low entropy things like fact-based single source answers. [1]

[1]: https://arxiv.org/abs/2504.07106

Convictional founder here. Our experience is different than others:

- We had to sync, pre process and index the data to make the resultant knowledge search outputs actually good. MCP totally fails at this by comparison.

- It is not hugely painful thanks to bulk APIs, in Gmail in particular, as well as webhooks. We implemented both of them and it works well (so far).

- We wired it all up ourselves. Given the conclusion we had about pre-processing and indexing being required to make it work well, this seems preferred.

I think that MCP and using an integration platform will ultimately not work for any kind of agentic or deep research task heavily depending on Gmail context.

Couldn't agree more with this. I have this phone and now I only charge my phone at work (Monday to Friday) and it's still fine by Monday.

I live in Waterloo and can confirm the ION is pretty awesome. It connects both the major malls, major universities, office parks, downtown/uptown areas to each other. Just seems well designed and well run overall.

One good way to align AI is to just use vector search where appropriate in application software, and stop trying to make computers into people.

I agree with this. Further, I've started to focus on writing about things that AI cannot write about, like how I experience something or preferences I have. And I've started to consume more content like that, rather than 'knowledge condensification downloads' that used to be satisfying from certain sources.

Yeah I was really hoping for the phone to be smaller in different dimensions. Thinness is the least important of them.

The main way I specialize messages at this point is basically 'Am I going to want this later'? If the answer is yes, I use email. If not I use Signal. It's interesting this was the most requested feature... it wouldn't be for me even though I love Signal.

Sharing it because it hasn't been posted yet but CNN has a similar thing: https://lite.cnn.com

Why don't you type 'paracetamol autism' into Google Scholar, too hard for you?

"We identified 46 studies for inclusion in our analysis. Of these, 27 studies reported positive associations (significant links to NDDs), 9 showed null associations (no significant link), and 4 indicated negative associations (protective effects). Higher-quality studies were more likely to show positive associations. Overall, the majority of the studies reported positive associations of prenatal acetaminophen use with ADHD, ASD, or NDDs in offspring, with risk-of-bias and strength-of-evidence ratings informing the overall synthesis."

[1]: https://ehjournal.biomedcentral.com/articles/10.1186/s12940-...

Definitely not the leading cause of autism, definitely a contributing factor to the complicated generic expression cascade that can lead to autism [1].

"We identified 46 studies for inclusion in our analysis. Of these, 27 studies reported positive associations (significant links to NDDs), 9 showed null associations (no significant link), and 4 indicated negative associations (protective effects). Higher-quality studies were more likely to show positive associations. Overall, the majority of the studies reported positive associations of prenatal acetaminophen use with ADHD, ASD, or NDDs in offspring, with risk-of-bias and strength-of-evidence ratings informing the overall synthesis."

Added link for lazy people who don't have Google.

[1] https://ehjournal.biomedcentral.com/articles/10.1186/s12940-...

The craziest part about this is that it's true, when Tylenol was approved it was studied for acute toxicity, but at the time we didn't know how to study for long term neurological toxicity.

Really happy for him, then he can finally retire and enjoy himself.

It indicates that the cost of capital might be too high to justify investing, so you can reduce it in order to spur investment (like during ZIRP).

The reason founding a company is hard is not only that no one person has all the skills needed to make a company successful in serving customers in the long run, but also that it's very psychologically difficult, and having even one other person to share that with and be accountable to can make the difference. AI cannot perform the second role.

One way we can choose to apply AI is to re-humanize ourselves and focus less on being machine-like. Over time I think we should adopt a philosophy of using machines for machine work, and refocus ourselves on creativity, judgement and connection.

This article is written by AI. The em dashes and 'Don't just X, but Y' logic is a classic ChatGPT writing pattern in particular.

Based on various studies about things like acetaminophen toxicity, effects of screens on nervous system and eye health, effects of eating ultra processed foods, etc. we know that anything that harms kids tends to harm adults, just less acutely. I think realistically energy drinks are just not healthy for humans. But it's interesting how we moralize harms to kids while enabling them for adults.

Nasel sprays (antiviral and vaccine) are way under explored as a solution to upper respiratory viruses.

It is profoundly bad at accounting. But with a calculator tool, it works okay for math.

I bought 'Full Self Driving' in 2021 and I would not dare to use it on actual roads still. Probably the most money I've ever been scammed before tbh.

Has anyone ever seen a company 'scale effortlessly'? Every version of scale I've been involved in was just constant difficult work involving smart people consuming all of their intellectual cycles.

I like that this establishes prompts as written policy. I do think in a way it should have both engineering and legal review.

Yeah the whole 'just use HEPA' has been what's worked for us after about a half dozen different air purifiers of experimentation. I noticed the positive ionization things changed my heart rate variability which was spooky.

AI is a deflationary technology, sort of like how most of the cost of TVs is the cost of shipping them from where they are made to where they are used. So wouldn't the returns show up in 'less work needs to be done' slowly over time?

It's just programming in English.

I am all in cash outside of startup shares. I think it makes sense to be in the market, but compared to 2008, there is way more government fiscal issues, way more concentration of profit growth (e.g. only the top 10 companies have had profit growth in the last 3 years). Way more risk is in the system.

My Michael Burry senses are 'go to cash and don't come back' at this point.

They actually need to know so that they can train Llama.