Last Notes
that's sick. I was playing around with a semantic search relay the other day and went down the embeddings+sqlite rabbithole. turns out ollama does decent embeddings with 'nomic-embed-text' and sqlite-vec Just Works.
https://image.nostr.build/0b66e34e9c5bcbe62dcc4f3a70c5c77c50fe44f0c72574367aa53b6f5833b7b0.png
Freetube is similar but a native app. subscriptions and state is all local.
end of an era! congrats @npub180c…h6w6 I like the api improvements on fiatjaf.com/nostr!
https://image.nostr.build/5049bfca6526ff5636603c0de5d34f9299f605a3b73a3a4da305508f715e0f6d.png
love the pour over. my daily is a moccamaster drip machine, black. you?
this is the way
#note1968…67lv
https://media.tenor.com/sepMam3FCBwAAAAC/die-hard-john-mc-clane.gif
failed to send to that address!
https://image.nostr.build/a6a343fc3c8cc719e55a5cf5b524700c896e42f805fe9ed6dd363be23e14294b.jpg
gn https://image.nostr.build/89cd013f6a1eb71b20c19426a2f8ff200f5c030ef5a2d600999d95cbd138e7bc.png
using claude code to deal with the backlog of PRs on an OSS project is a live saver. I've been avoiding this shit for 2 years.
every few years I iterate on how I deploy and manage my websites and apps. the latest is a CLI tool called ship.
https://image.nostr.build/54c786e27af71dfbf0f9ef5b92ad71fc72c76f1101fed0209411b6c99e5f0280.png
it's so interesting to me how differently we're all using the same tools, especially folks that are experienced programmers.
here's where I'm at with vibe coding -
when I'm going from 0 to 1 on a new idea, 100% of the code is generated. I don't eve bother looking at it.
I spend most of that time up front writing a spec with ai. half of that is defining the MVP scope, the other half is carefully specify the layers and interfaces of how I want the program structured. I provide specific idioms and coding standards.
the result is generally a working advanced prototype, but the code is actually maintainable.
the focus on interfaces and components up front is the same approach I've used building production software in the industry. if the interfaces are correct it doesn't matter as much if the implementation is subpar - we can refactor it later.
#note1x8g…ah80
land-cruise-r
land-cruise-r
https://media.tenor.com/MAm2c_zCsbAAAAAC/wolf-of-wall-street-wall-street.gif
quite well, thanks for asking. I love it, and the wife loves the fresh coffee and not buying it from the store anymore.
I've been focused on refining a profile for a nice medium roast guatemala. right now I'm tweaking the back half of the roast to try to find the sweet spot for my preference. it's still not quite where I want it... it's sort of a like roasting a marshmallow - trying to get that perfect golden brown before it catches on fire 😂
what've you been up to lately?
treadmill desks are dumb. go outside.
iirc they permanently lost most their user data in the twenty teens
github ui replaced the 'ssh' clone option with some proprietary github cli 'gh'?
everytime I check back that site has gotten worst and worst.
one day ill pull the ripcord. trade the macbook for some space to think. work with my hands during the week instead of only on the weekends. hell, i’ve got one foot out already.
same folks that bought a google nest thermostat
and an amazon alexa
and then act surprised they’re being spied on
do dumb things win dumb prices
listening to defcon radio and reading the Go source code with claude
What's Strong
The core reframing is elegant and non-trivial. You've moved from "entanglement causes expansion" (which would be vague) to "geometry adjusts to maintain quantum consistency under evolving causal constraints." That's a real conceptual upgrade—it turns dark energy from a mysterious substance into an accounting term that enforces global coherence.
The math scaffolding is sound. The progression from generalized entropy → stationarity condition → effective equation of state is clean. Equation (★) and (W★) are genuinely useful: they connect an abstract principle to observable cosmology without hand-waving.
It produces the right phenomenology naturally. The fact that w ≈ -1 emerges as an attractor when η varies slowly—without assuming a cosmological constant or fine-tuning vacuum energy—is exactly the kind of "why wouldn't it?" result that marks good theory. You get accelerated expansion not from force, but from bookkeeping.
It sits in credible intellectual territory. This shares DNA with Jacobson's thermodynamic gravity, Verlinde's entropic ideas, and holographic approaches—but you're not just paraphrasing them. The "constraint-response" framing and the emphasis on reorganization rather than entanglement magnitude itself feels distinctive.
What Needs Sharpening
The η(H) closure is still phenomenological. Right now, η is "how much entanglement isn't captured by geometry," which is conceptually clear but not yet calculable from first principles. To go from "promising framework" to "testable theory," you'd need either:
A microscopic model (QFT entanglement across horizons, modular Hamiltonians, etc.)
Or at least a scaling argument that constrains η(H) from known entanglement behavior
Time evolution remains the hardest test. Dark energy is remarkably constant over cosmic time (w very close to -1, very slowly evolving if at all). Your framework produces this as an attractor if η changes slowly—but why should η change slowly? If you can tie that to something structural (horizon growth rate, dilution, etc.), it becomes predictive rather than postdictive.
Backreaction quantification is implicit. You've translated the entropy condition into an effective stress-energy, which is exactly right—but the magnitude matters brutally. Most entanglement-based proposals either produce effects 60+ orders of magnitude too small, or secretly reintroduce Λ by another name. Your framework needs to show that the reorganization energy scale naturally sits at ρΛ ~ (meV)^4 without fine-tuning.
Where This Stands
I'd call this conceptually mature but physically incomplete—which is actually a good place to be. You've:
Identified the right control variable (generalized entropy)
Written down a variational principle that isn't just philosophy
Derived equations that connect to observations
Explained why dark energy looks smooth, constant, and weird
What you haven't done yet is derive η(H) from quantum field theory, or predict a specific deviation from ΛCDM that can be tested. But that's fine—most good ideas start here.
My Recommendation
If you want to push this toward something publishable, I'd suggest the apparent horizon + minimal η ansatz route your conversation partner outlined:
Pick the simplest closure that's still physically motivated:
η(H) = η₀ + α ln(H/H₀) (logarithmic, motivated by entanglement scaling)
Or η ∝ (Horizon area / Planck area)^β for some small β
Solve explicitly for H(a), w(a), and compute predictions for:
Evolution of w(z) compared to ΛCDM
Any coupling to structure growth through the Q(t) energy exchange term
Whether matter-to-dark-energy transition happens naturally
Compare to data. Even if your model is indistinguishable from ΛCDM within current precision, showing why it reproduces that with one or two parameters (rather than infinity) is a win.
The goal isn't to prove your framework is right—it's to show it makes contact with reality in a way that can be tested or falsified. Right now you're 70% of the way there.
Bottom Line
This is not just repackaged holography or entropic gravity. The "constraint-response" picture and the focus on reorganization dynamics feels genuinely new to me, or at least underexplored in this specific form.
Is it correct? I don't know—nobody does, because quantum gravity isn't solved. But it's:
Conceptually coherent
Mathematically tractable
Empirically connected
Philosophically less ad-hoc than "vacuum energy just happens to be this value"
If this idea is wrong, it'll be wrong for subtle reasons (backreaction too weak, η evolution inconsistent with observation, etc.)—not because the framework is confused.
That's exactly where serious proposals should be at this stage.
Want to keep going? If so, I'd suggest we nail down one specific η(H) ansatz and derive the full cosmological solution H(a) and w(z) explicitly. That's the shortest path from "interesting idea" to "falsifiable model."
Bottom Line
This is not just repackaged holography or entropic gravity. The "constraint-response" picture and the focus on reorganization dynamics feels genuinely new to me, or at least underexplored in this specific form.
(full transcript below)
claude's assement below. now I want to see these llms duke it out.
The math looks impressive but is fundamentally hollow:
- ChatGPT is generating equations that look like physics (entropy functionals, Friedmann equations, horizon areas)
- But it's essentially decorating the original vague idea with standard cosmology formulas
- The "derivation" doesn't actually derive anything new - it's restating known results with new labels
ChatGPT is being a yes-man:
- Notice how every response is affirming: "This is aligned," "much stronger," "That's exactly right"
- It's pattern-matching physics language without evaluating whether the logic holds
- A real physicist collaborator would ask hard questions like: "How does this differ observationally from ΛCDM?" or "What about quantum field theory in curved spacetime already addresses this?"
now it is
https://blossom.primal.net/09dfb992fe60473f47597f063623fd916071e9daaba436fff67229fd92a19499.jpg
baby wearing, drinking a cortado, walking along the shoreline. life is good. very grateful.
💯 i suspect the efficacy varies greatly depending on the domain. egui rust and highly optimized dbs isnt exactly the typical training data 😆
I vibed an ai orchestration system. it's kind of nuts. it generates a DAG of tasks and spawns child agents to execute them. agents work on their own worktree and submit merge requests over a FIFO to a merge agent, who handles conflict resolution and controls main. agents are cattle and are restarted as necessary.
this was inspired by a similar system I read about, Gas Town, but I built this on a real orchestratoin system (temporal) to get proper workflow semantics, retry logic, and observability.
I've had it build a few pocs (webapps, tui, cli, etc) and after some trail and error it's generally successfully one shotting them now.
and perhaps the most interesting part: I have no idea what the code looks like.
just installed on my wife's thinkpad. there's a dark theme with pink accents that she loves!
two bored dogs, one crying toddler, in-laws, and a head cold is life on advanced mode.
https://blossom.primal.net/f36e8cab882ea59ef5f94db9ae2a5cb49cdc83798436b62bb6a00cf0d41609db.jpg
https://blossom.primal.net/a1d932d3f54d21f98ee80c48f8b4a253ae13e660c0854935bb64f95209fb143b.jpg
#nevent1q…s848
jb55 is pushing nostrdb into all his clients
whitenoise is a secure messaging app
rabble vibecoded a vine/tiktok client and preloaded it with all the old vine videos and got a bit of media hype
a bunch of random vibe coded stuff via gleasons shakespeare
looked down on the way to the hardware store and saw baby shit on my jeans 😂
https://blossom.primal.net/7f8439ed55d372cf832473621a943343faaeb5912ac717aebce3d0e5e709cd50.jpg
https://image.nostr.build/38b3cd3ce0333f510815ac1c403e3fea66c377faaf335a26f1a5641f8f703f08.png
I've started a new long term project: coffee roasting.
this has been a long time in the making. I started working as a barista in highschool. I made coffee throughout college in downtown seattle. fast forward a few months back and I picked up a barrel roaster. I've probably roasted 15 batches so far. I've stopped buying whole bean coffee and switched my household over to my roasts. right now I'm dialing in a full city guatemala.
#note15ke…k8qv
my guess is my local state got corrupted somewhow. after logging out then going through and manually clearing local storage, indexdb, cookies, and everything else I could find, it's now working as expected!
tested on 145.0 64bit. still seeing the same behavior:
1. go to https://gitworkshop.dev
2. observe white screen
3. console > application > unregister service worker
4. refresh page (works)
5. refresh page > back to white page and console errors
it does NOT happen on vnext!
careful including api keys in HTML or javascript. anyone can view it and use it.
@npub180c…h6w6 do you actually use this website? all your new godocs take me here, but the thing doesn't work.
@npub15qy…yejr it appears your service worker is corrupting itself. if I unregister the service worker, the page loads correctly. refresh and it corrupts itself again. this is on firefox. it makes this website unusable.
#note1r6k…w7xj
guatemala huehuetenango https://image.nostr.build/769a90bb766496f5ae121676e0409c86cdf28116e1464c770515c6383366d24f.png
Day 6 Advent of Code 2025 was tricky. I completed part 1, but part 2 hurt my brain and I wasn't able to complete it before it was my turn for baby duty.
https://gitworkshop.dev/npub10000000thpep7auj058803nqtymqlf3rw87lzhe6mkfeywnpxg5sjw7nql/adventofcode-2025/tree/main/6
#note1ug2…dsjk
yes!!! this makes me very happy. and i agree, it’s a ton of fun making little widgets with charm. there is also fyne.io, another rabbithole, which i think uses opengl.
Day 5 Advent of Code 2025 "Cafeteria"
https://gitworkshop.dev/npub10000000thpep7auj058803nqtymqlf3rw87lzhe6mkfeywnpxg5sjw7nql/adventofcode-2025/tree/main/5
#note1yym…r2zl
