Matt is amazing. After checking out his compiler optimizations, maybe check out the recent interview I did with him.
What I’ve come to believe is this: you should work at a level of abstraction you’re comfortable with, but you should also understand the layer beneath it.
If you’re a C programmer, you should have some idea of how the C runtime works, and how it interacts with the operating system. You don’t need every detail, but you need enough to know what’s going on when something breaks. Because one day printf won’t work, and if the layer below is a total mystery, you won’t even know where to start looking.
So: know one layer well, have working knowledge of the layer under it, and, most importantly, be aware of the shape of the layer below that.
The “understand one layer below where you work” is something my professors at uni told us 10+ years ago. Not sure where that originated from, but I really think that benefited me in my career. I.e understanding the JVM when dealing with Java helped optimize code in a relatively heavyweight medical software package.
And also, it’s just fun to understand the lower layers.
I did find beads helpful for some of this multi-context window tasks. It sounds a little like there is some convergence between what they are suggesting and how it give you light weight sub tasks that survive a /clear.
> It sounds a little like there is some convergence between what they are suggesting and how it give you light weight sub tasks that survive a /clear.
I do see the convergence there. Beads gives you that "state that survives `/clear`," and Anthropic’s harness tries to do something similar at a higher level.
I've been thinking about this with a pretty simple, old-school analogy:
You're at a shop with solid engineering and ticketing practices. You just hired a great junior developer. They know the stack, maybe even the domain basics, but they don't yet know:
- Your business processes
- The quirks of your microservices
- Local naming conventions, standards, etc.
- Team norms around testing, logging, and observability
You trust them with important tasks, but expect their context will frequently get blown away by interruptions, meetings, task-switching, and long weekends. T handle this, need to make sure each ticket or note contains enough structured info so that when they inevitably lose context, they can pick right back up.
For each ticket, you'd likely include:
- Personas and user goals
- Acceptance criteria, Given/When/Then scenarios
- Links to specs, documentation, related tickets, or prior art
- A short summary of their current understanding
- Rough plan (steps, what's done/not done)
- Decisions made and their rationale ("I chose X because Y")
- Open questions or known gotchas
End of day Friday, that junior would ideally leave notes that answer:
"If I have total amnesia next Tuesday, what's the minimum needed to quickly reload my context?"
To me, agent harnesses like Anthropic's or Beads are just formalizing exactly this pattern:
- `/clear` or `/new` is like a "long weekend brain wipe."
- Persistent subtasks or controllers become structured scaffolding.
- The crucial piece isn't remembering everything, just clearly capturing intent, decisions, rationale, and immediate next steps.
My confusion about Anthropic’s approach is why they're doing this over plain text files or JSON, instead of leveraging decades of existing tracker and project-management tooling—which already encode this exact workflow and best practice.
My 2 cents on BQN: I am certainly a novice with array languages, but I know they have conceptual power.
Looking for a modern, powerful language centered on Ken Iverson's array programming paradigm
BQN aims to remove irregular and burdensome aspects of the APL tradition, and put the great ideas on a firmer footing.
And BQN seems like the closest thing to a 'modern' array language. Modern, meaning, looking like my biased version of what language should look like.
Open source, has namespaces, and you can define your own operators and so on.
I actually previously used Jekyll! Built this largely because I want full React component functionality sometimes. Also I think Jelyll gave me some issues with routing that I didn’t like.
> In the early days of FedEx, Smith had to go to great lengths to keep the company afloat. In one instance, after a crucial business loan was denied, he took the company's last $5,000 to Las Vegas and won $27,000 gambling on blackjack to cover the company's $24,000 fuel bill.
Some who take on unreasonable risk will be among the most successful people alive. Most will lose eventually, long before you hear about them if they keep too many taking crazy risks.
Who is a great genius, and is who is just winning at "The Martingale entrepreneurial strategy"?
You know, it only just now occurs to me to wonder if the blackjack story is the public sanitized version of "how I got $24k because I'm not allowed to tell you the real version"
What this version of the FedEx story doesn't mention is that Fred was already stiffing his pilots on their salaries. Taking the last money in the company and deciding that the best use for it was the blackjack table in Vegas and not paying his employees ... worked well, but it was a gamble, let's be clear, not a calculated decision - like you say, not the decision of a "great genius". It goes a different way, and you have "FedEx founder decides to go gambling, leaving his employees without paychecks".
The casino is an extremely rational savings model if you expected to constantly be robbed and want to convert small (and thus not worth robbing) income streams into occasionally large sums of money to be spent rapidly. I.e. say you are a north korean worker in China/Russia and you occasionally get small change to spend on cigarettes, you could gamble it every 'paycheck' and eventually buy a phone to escape with the winnings.
Filipinos have a more predictable low-loss version of this call Paluwagan.
It can also be a way to evade capital controls. You know you'll lose, but otoh you also know you'll probably not lose _everything_ and you buy in with RMB and cash out in HKD.
> simultaneously extremely similar to and extremely different
yeah, I don't understand the change tbh.
It's said Eric Heisserer spent years and years on the screenplay so I'm assuming he couldn't sell the original version. But it's a bit like making fight club and removing the big reveal. It ends up feeling the same, but not having the same impact and meaning almost the opposite.
> The best reason to take multiple life extension supplements is to hedge our bets, because we really don’t know which of them are effective in humans.
And earlier:
> Personally, I take large doses of rapamycin 2 days a week, 8 weeks per year. For personalized recommendations, you can consult your favorite life extension doc.
I recently saw a patient with overwhelming MRSA sepsis with multiple foci of infection including epidural abscess (around the spinal cord), and meningitis. This person was taking rapamycin presumably for "life extension" purposes. Almost certainly the immunosuppression from the rapamycin made the infection much worse.
I'd be very wary of taking an immunosuppressive drug as an otherwise healthy person for theoretical life extension properties.
Rapamycin modulates the immune system. I get that he's probably consulting a doctor but can you imagine taking this risk during a pandemic or even in older age? It makes me uncomfortable to play around with these very powerful drugs.
The dosage for longevity is supposed to be low enough that this risk is minimized. Lots of things you do modulate your immune system (including e.g. exercise). It's a risk/reward thing, every time you get into your car you're also taking a longevity risk.
I think there are some proper human trials happening but the jury is still out.
People are still getting nerve damage from too much vitamin B6 in energy drinks and vitamin supplements, and that's a well known and widely taken vitamin. The idea that you can take experimental drugs your entire life at little risk is optimistic.
Surprisingly, many people seem to think that pushing a few random pills into a machine optimized over some million years of evolution will tune it so it works better. Go figure...
Yes, although even for modern medicine curative and preventative strategies are very distinct. Sure, they'll give you pills to compensate for a problem you already have. But there are few meds that protect you against stuff you'll maybe catch in the future. Vaccines and antibiotics are obvious examples, but I'm not aware of many others. The rest of preventative strategies overwhelmingly consists in correcting deficits or excesses (calories, vitamins, sleep, exercise etc.)
Never put 100% of your savings into a single slot machine. Take 10% of your savings to 10 different casinos and distribute them to 10 slot machines in each, in order to hedge.
Oh there's plenty of people selling "side effect free" life extension supplements. But there's another name for side effect free medication: effect free.
Not doubting all this, but the possum thing is interesting.
They were in southern ontario in my youth in essex county ( late 80s ). And google says they were reports as far back as the 60s of scattered sightings.
Also this article in acmqueue by Matt is not new at all, but super great introduction to these types of optimizations.
https://queue.acm.org/detail.cfm?id=3372264
reply