If this source [0] is true then 65% of audiobooks (in 2022) were non-fiction. Likewise that the audiobook industry has grown by nearly 3x since 2022. So, by my math, it's simply that people prefer to listen to self-help books (which matches my own experience).

[0] - https://electroiq.com/stats/audiobook-statistics/

Normally, I work on my core work plus something tangentially related (e.g. 20% projects). The 20% projects keep my attention while the core work is LLM'ing.

So far every Grep vs RAG discussion I've seen conflates overlapping factors. The most common is simply that a company rebuilt their pipeline from scratch and fixed a bunch of problems. The worst is when they go from one-shot RAG to multi-step Grep and completely miss the fact that multi-step RAG would likely get them similar results.

At the end of the day, the most important thing is knowing the _product features_ your users care about and making sure that's represented in the pipeline.

It's really the only way for companies to survive and go up market, unfortunately.

It’s almost like you’re offering a taxi service to people who want to own cars. Now, here’s the thing, even when people own cars, they still outsource driving in certain situations: taxi home from the bar, limo/party bus for evens, rides to the airport, group tours, long distance, etc, etc.

Instead, of general, focus in on a specific use case and make it as simple as possible to get good results with that specific use case.

Even, then, I’m doubtful of the ability to get traction in this space. People don’t really appreciate AI generated art. The only place obviously AI generated art seems to get traction is terrible FB ads and NSFW content.

I was in a similar situation where Claude finally helped me make a breakthrough. It seems that my issue is entirely related to histamine levels.

* H1 (like Zyrtec) help block the body’s response to high histamines

* H2 (like Pepcid) actually worsen my symptoms overtime because my body was readjusting baseline

* Quercitin and DAO enzyme help massively.

Also being extremely mindful of histamine levels in food helps keep my baselines low. This can be tricky because histamines collect as food ages (so two batches of something can lead to different responses) and some “low histamine” foods can still trigger your body to dump histamines (so they actually act like high histamine foods).

I tried all sorts of different things, but couldn’t nail down a pattern for years. Even worse, some days I could tolerate something and others it would blow me up. It wasn’t until I started telling Claude literally everything I consumed and how I felt that it dialed into histamine response triggering mast cell activation.

I started to realize that a lot of “low histamine” foods either build up histamines as they age (so leftovers kill me) and some “low histamine” foods actually cause the body to naturally liberate histamine (essentially mimicking high histamine foods).

It’s extremely hard to get a diagnosis because the triggers are seemingly random and don’t always correlate with common buckets.

You loose sooooooo much meaningful context and information when you transform something into a knowledge graph. Simple cases like "Gabe is CEO of Valve" map nicely to a graph, but things like "Matt Garman is CEO of AWS" don't represent that AWS is a sub-company of Amazon (with it's own CEO).

Additionally, one of my biggest gripes of Claude's memories and every memory system I've worked with is they completely fail to capture intent. The architecture notes I documented while doing a wild spike on a critical infrastructure component absolutely should not be referenced in every day work. Yet, somehow, that type of memory always works it's way into unrelated sessions.

Just missing some em-dashes.

I switched to HCL and tolerate it massively better.

The HCL form is more expensive, but does wonders for me. None of the negative side effects.

I need more mechanisms for controlling long-running sessions and dynamically injecting my thoughts, correction, and nudges rather than faster ways to burn through my tokens without knowing if the results are going to be correct.

Mechanical refactors are relatively straight forward for agents.

I've actually intentionally switched back to 4.5. I hated 4.7 so much that I decided to jump back all the way to 4.5.

Now that I've been using 4.5 for a few weeks, I find it significantly more reliable but a bit more forgetful than 4.6/4.7. I'm okay with that because it's really easy to identify this forgetfulness and nudge it.

I found 4.7's adaptive thinking to be extremely unreliable. It seems to overcorrect on the current message without considering the difficult of the overall problem. I wonder if 4.8 will improve on that.

Corporations are an artificial entity that literally anyone can make. Even things like property ownership are somewhat artificial. Lots can generally be split and joined through a process.

This allowance of artificial entities voting seems to open a rabbit hole of secondary issues.

* Start an LLC/C-corp for a trivial amount of money.

* Purchase land, but instead of paying with it via a personal check, you need a touch of foresight so you can "capitalize" the corporation you just started. Write the check from the corporation, instead of your personal checkbook.