b0a04gl

why tf is git still running submodule hooks during clone at all. like think. youre cloning a repo which you didnt write it or audit it. and git just... runs a post checkout hook from a submodule it just fetched off the internet. even with this CRLF bug fixed, thats still bananas

what if we can make these diagrams synchronized with reality. you need the diagram to pull from the same source of truth as your actual infrastructure - whether that's terraform state, kubernetes manifests, or service discovery. that way diagrams become less historical artifacts and more of living documentation

this is another layer of abstraction on top of an already broken system. you're running html through an llm to get markdown that gets rendered in a terminal browser. that's like... three format conversions just to read text. the original web had simple html that was readable in any terminal browser already. now they arent designed as documents anymore but rather designed as applications that happen to deliver some content as a side effect

before this you had to trust that claude would follow your readme instructions about running linters or tests. hit and miss at best. now its deterministic. pre hook blocks bad actions post hook validates results.

hooks let you build workflows where multiple agents can hand off work safely. one agent writes code another reviews it another deploys it. each step gated by verification hooks.

every developer knows safari is the new ie6 but we all just shrug and build native apps anyway because what else are you gonna do. leave 50% of your users on the table. classic embrace extend extinguish but in reverse. embrace web standards then purposely not extend them so you can extinguish competition

10kloc for command line parsing. TEN THOUSAND LINES. pico-args does it in 700 lines and probably handles 99% of real world use cases. compile times go to shit binary size bloats and for some edge case you'll never hit.most CLI tools need what three four flags max, maybe a subcommand or two. you don't need the swiss army knife of argument parsing for that. tried replacing clap with pico-args on three different projects last month. 80% reduction in compile time every single time. binary went from 8mb to 2mb on one of them.the "disk space is cheap" argument's acceptable partially but compile time isn't. developer experience isn't. startup time isn't. memory usage isn't

in async code ,errors belong to the task ,not the caller.

in sync code ,the caller owns the stack ,so it makes sense they own the error. but async splits that. now each async function runs like a background job. that job should handle its own failure =retry ,fallback ,log because the caller usually cant do much anyway.

write async blocks like isolated tasks. contain errors inside unless the caller has a real decision to make. global error handler picks up the rest

why is bot detection even happening at render time instead of request time. why can't tell you’re a bot from your headers, UA, IP, TLS fingerprint. imo making it a surveillance. 'you're a bot, ok not just go away, let’s fingerprint your GPU and assign you a behavioral risk score anyway'

provenance model basically turns memory back into a typed value. finally malloc wont just be a dumb number generator, it'll act more like a capability issuer. and access is not 'is this address in range' anymore, but “does this pointer have valid provenance”. way more deterministic, decouples gcc -wall

how deterministic is the emit really. if i feed same expression tree twice,same node layout same captures. do i get exact same bytes out every time (ignoring reloc) or not. if output produced is byte stable across runs for same input graph ,that opens up memoized JIT paths.worth checking if current impl already does this or needs a pass to normalise alloc order

why do fingerprinting always happens right at connection start ,usually gives clean metadata during tcp syn. but what is it for components like static proxies or load balancers or mobile networks ,all of these can shift stack behavior midstream. this can make this activity itself a obsolete

what if this wasnt something you add after infra but the checkpoint you start with. right now you spin up a vm or db then wrap vpn or firewall around it. but imagine writing access rules first in way : 'team ml can hit service x' or 'web app can hit this backend' and the system wires infra from that.. infra becomes a side effect of access intent. access isnt something you cant guard always( as things move fast, breaks fast), it's may become seed where you can design with.

Mitre ATT&CK's T1497.001 (VM Detection) lists SMBIOS checks as a known vector means its open for injection anyways.

i did one little expirement on faking VM's powersupply. done it with 'HotReplaceable=Yes' and 'Status=OK', and you suddenly look like a $5k baremetal server.

cmd used

pip install dmigen dmigen -o smbios.bin \

--type0 vendor="American Megatrends",version="F.1" \

--type1 manufacturer="Dell Inc.",product="PowerEdge T630" \

--type39 name="PSU1",location="Bay 1",status=3,hotreplaceable=1

ipv6 only machine still reaches ipv4 sites because dns64 upstream is just faking AAAA records ,makes it look like everything is native ipv6. this part of the trick is happening somewhere else which's not controllable. if dns64 breaks or stops doing the mapping properly then this might break

found one more flakiness over cross platform, when seed mt19937 same way on linux and windows, same compiler, same code... but problem is std::random_device or libc internals differ under the hood. some platforms do random_device as true hardware entropy, others fake it or seed from diff system sources. so seed retrieved isn't stable cross platform. that means mt19937 starts from diff states, causing different random sequences

it's not a bug in mt19937 itself, it's how random_device (or libc randomness) works differently across environments. makes cross platform tests flaky even when logic is rock solid

>

std::random_device rd; // might differ per platform

std::mt19937 gen(rd()); // seed depends on rd output

std::uniform_int_distribution<> dist(1, 100);

int random_number = dist(gen); // different on linux vs windows tho same code

no way this delay's about gpus lol. deepseek prob has r2 cooked already. r1‑0528 already pumped expectations too high. if r2 lands flat ppl start doubting.

or

who knows maybe they just chillin watching how west labs burn gpu money, let eval metas shift. then drop r2 when oai/claude trust graph dips a bit

this works like functional programming where every symbol is a pure value and operations compose into clean, traceable flows. when you hit an ambiguous step, the model steps in. just like IO in FP, the generative call is treated as a scoped side effect. this can engage your reasoning graph stays deterministic by default and only defers to the model when needed. crazy demo though, love it

file layout is the interface here lol you can literally walk the pipeline.. lexer parser codegen linker all just sit where they should. the dir was the flow. back then structure = filesystem. we can cd trace src to bin just by lookin at folders

when a route doesn't come back as a roundtrip , like you fly LHR > HKG but not the return . how does that usually get handled on your end? do you deadhead back, get reassigned regionally or wait out a layover cycle?

they exist because whole language built to treat expressions as firstclass citizens : blocks, ifs, matches, even macros as expressions that return values. so once you internalize that, all these weirdo one liners are artifacts. just artifact of a system where expressions compose infinitely. the syntax tree runs deeper than most people's habbits allow. you hit that depth and brain says this is wrong but compiler's allowing.

image gets compressed into 256 tokens before language model sees it. ask it to add a hat and it redraws the whole face; because objects aren't stored as separate things. there's no persistent bear in memory. it all lives inside one fused latent soup, they're fresh samples under new constraints. every prompt tweak rebalances the whole embedding. that's why even small changes ripple across the image. i notice it like single shot scene synthesis, which is good for diff usecases

these layouts break kerning rules. render engines expect horizontal flow, steady spacing. but with same sizer or echoed lines, glyph logic goes off path. spacing's no longer font native, it's forced by layout. font stops being just visual, becomes part of layout logic. whole engine ends up doing things it wasn't ment for. then layout will start mutates typography logic iteslf

flows make sense here not just for size but cuz they're fully invertible and deterministic. imagine running same gen on 3 iphones, same output. means apple can kinda ensure same input gives same output across devices, chips, runs. no weird variance or sampling noise. good for caching, testing, user trust all that. fits apple's whole determinism dna and more of predictable gen at scale

seen this pattern a before too. faker holds shape without flow. real tables come from actions : retry, decline, manual review, all that. you just set col types, you might miss why the row even happened. gen needs to simulate behavior, not format

how're you handling latency on turn overlaps : buffered stream with early intent cutoff or full duplex with partial decoding?

supertokens did the same thing from bengaluru. didn’t start loud. just showed up with clean abstractions that didn’t leak. you could tell someone had wrestled with real auth mess before touching a single line. it worked, across teams, stacks, workflows

better auth gives off the same shape. that gets well adopted because it survives scaling without needing a rewrite

same pattern and diff origin place. someone holding the whole stack in their head long enough to ship something

lets say if i someone wants to keep using bittorrent dht for peer finding but swap out quic for something else maybe grpc, does the lib support that split clean? asking from a modular embed first tooling pov, where discovery logic needs to outlive or outswap transport depending on deployment

i remember koramangala, 5th block specifically, mid 2023. that blue tokai outlet next to roastery was ground zero. half of early stage bangalore was working from there. two pm to six pm you'd overhear: investor calls, pitch deck review, even product teardown with some YC alum. no seats inside so i parked at the outside bench near the window. wifi barely reached there. next to me this guy's debugging something on a steamdeck looking devkit. i half glance over and ask if it's AWS creds, he goes 'nah, it's some edge TPU , google keeps timing out cold starts'. we start chatting.

turns out he's building vision for offline-first retail. he's got no frontend, just a python backend. i scribble something on a napkin about fast-booting wasm modules from disk cache. 3 weeks later he pings me on telegram saying they got boot time down from 14s to 2.8s using a variant of that.

never met him again. never even learned his startup's name. but that entire bottleneck cleared because two people overheard a swear word near a bad socket.

we maynot recreate that on a discord channel. there's no incentive to overshare when you're not spatially co-located. bangalore 2023 worked because entropy was high and friction was low

i see one crazy real leverage: every prototype built here is a frozen snapshot of someone's product thinking in motion. like u can literally watch how ppl prioritize flows, kill features mid-wireframe, choose friction over flexibility. it’s raw cognitive output

if lovable ever starts versioning those moves, storing reasoning behind edits, even lightly, u got a time-series of product intuition across thousands of users. that’s applied decision memory.

there's a window here to become the place where product sense gets archived and replayed.

it's more to do with the obvious economic layer underneath. you give a format new life only if there's tooling and distribution muscle behind it. adobe, apple, chrome, ffmpeg etc may not get aligned at the same time. someone somewhere wants apng/hdr/png to be a standard pipe again for creative chains; maybe because video formats are too bulky for microinteraction or maybe because svg is too unsafe in sandboxed renderers. and think onboarding of animations, embedded previews, rich avatars, system wide thumbs ; all without shipping a separate codec or runtime. every time a 'dead' format comes back, it's usually because someone needed a way around a gate