HackerTrans
TopNewTrendsCommentsPastAskShowJobs

ignoramous

18,619 karmajoined há 15 anos
mz at celzero dot com

https://rethinkdns.com/

follow:

sec: tptacek, moxie, nickpsecurity, strcat, agl, SwellJoe, drewcrawford, schoen, mirimir, secfirstmd, mjg59, userbinator, gorhill, rgovostes, lallysingh, malandrew, mikewest, jedberg, wtarreau, michaelaiello, segmondy, whitequark_, jsnell, salgernon, geofft, jlund, sinak, alecmuffett, mrb, nneonneo, pwg, viraptor, indutny, jart, Danenania, cyphar, lisper, drfuchs, mahmoudimus, lrvick, woodruffw, lvh, FiloSottile, pgl, mh_, amarbi, psanford, jedisct1, lotharrr, axoltl, cperciva, sleevi, kwantam, syncsynchalt, Moral_, saurik, cryptonector, JoachimS, j2kun, woodruffw, bjackman, nneonneo

start: pg, sama, garry, mbesto, coffeemug, davidu, rms, xal, malgorithms, Alex3917, jacquesm, jl, AndrewWarner, emmett, ig1, anateus, lpolovets, ivankirigin, sahillavingia, Joshua, ericflo, immad, rdl, joshfraser,s gdb, grellas, gatsby, mayop100, bryanh, josephsunny, ayw, pbiggar, sytse, csallen, joshu, jhuckesteinm, pclark, whockey, sjtgraham, jenthoven, aresant, mayop100, cmdrtaco, zt, erohead, tomhoward, drderidder, rfrey, pauldix, akcreek, mojombo, salsakran, armon, drob, pavlov, goodmachine, snowmaker, sorenbs, simonw, zds, jessepollak, raviparikh, buf, paraschopra, ahaseeb

prog: aphyr, KiranDave, peterwaller, saosebastiao, amirmc, lhorie, judofyr, nikita, huhtenberg, pron, Animats, dmbaggett, chris_wot, aaronbrethorst, grey-area, drewg123, dom96, jashkenas, rich_harris, jordwalke, Homunculiheaded, mark_l_watson, kibwen, Sir_Cmpwn, KenoFischer, ahoyhere, scrollaway, trishume, dbaupp, skrebbel, cryptica, peterhunt, rauchg, TazeTSchnitzel, syrusakbary, megous, jemfinch, fogus, jcelerier, 1st1, anderskaseorg, willvarfar, Veedrac, kodablah, mraleph, enriquto, joosters, loeg, jorangreef, ot, asicsp, felixhandte, ezmobius, r1ch, luckydude, btilly, dherman, Manishearth, carllerche, gregdoesit, q3k, codahale, kuschku, _vbdg, enneff, phiresky, mitchellh, JoshTriplett, NovaX, kmavm, nullc, EvanYou, colanderman, sadiq, sereja, ot, pfdietz, c-smile, xena, WalterBright, izacus, bradfitz, scott_s, pizlonator, pojntfx, lifthrasiir, kentonv, barsonme, rspivak, tekknolagi, ndesaulniers

systems: brendangregg, DannyBee, steveklabnik, fsk, pcwalton, jbk, ajross, yosefk, netguy, minimax, munificent, ColinWright, beat, zwischenzug, derefr, jandrewrogers, shykes, lallysingh, dochtman, SamReidHughes, hnkimb3558, rurban, gonzo, bogomipz, marknadal, mazieres, bramcohen, koverstreet, bkanber, mafintosh, ksec, amluto, kyledrake, mrkurt, dsl, dspillett, kevinburke, catwell, kmod, scarface74, eru, sanxiyn, znpy, influx, ncmncm, toast0, evil-olive, bonzini, monocasa, ncopa, bfirsh, jefftk, tlrobinson, kiwicopple, stavros, alexellisuk, hardwaresofton, jhgg, jeffbee, hinkley, danielbmarkham, cpuguy83, silverstorm, josephg, tialaramex, apenwarr, nocarrier, KMag, kllrnohj, astrange, bayindirh, inkyoto, fweimer, mananaysiempre, antics

mods: dang

bio: Fomite, comicjk, anderspitman, danieltillett, wgrover

linux: rwmj, pdkl95, linuxlizard, polvi, phillips, zozbot234, wmf, beagle3, megous, bcrl, cdesai, lgierth, monocasa, loeg, swetland, jchw, abarth, microcolonel

misc: phkahler, davidw, antirez, gwern, patio11, jgrahamc, darksaints, jamwt, nostrademons, plinkplonk, mikekchar, holman, mikeash, edw519, jrockway, noonespecial, staunch, petercooper, jmathai, tzs, jacques_chester, coldtea, peteretep, happy-go-lucky, aaronbrethorst, mtgx, codingdave, jawns, jvns, CPLX, tomcam, ethomson, HenryR, JoeAltmaier, jerf, pjc50, dmix, jcr, alexbowe, JulianMorrison, adamnemecek, capnrefsmmat, detaro, calinet6, dargonwriter, tootie, kqr, comex, eloff, andersource, dantiberian, lern_to_spel, swyx, pgeorgi, sago, zokier

os: vezzy-fnord, rbehrends, vardump, amirmc, pjmlp, rsync, waddlesplash, eyberg, penberg, ambrop7, shuss, amscanne, tytso, surajrmal, captainmuon, Morgawr, quotemstr, olliej, kllrnohj, pizlonator, tadfisher, faragon

db: craigkerstiens, teraflop, ifcologne, espeed, gopalv, thekozmo, lorenzhs, SQLite, mytherin, pgaddict, sumeer, NovaX, arjunnarayan, dmoura, eatonphil, cube2222, zX41ZdbW, benbjohnson, JoelJacobson, benesch

graphics: pcolton, Jasper_, macawfish, kvark, Agentlien, shmerl, Atrix256, jasondavies, bhickey

net: zx2c4, keithwinstein, bsder, walrus01, apenwarr, newman314, stuntprogrammer, bluejekyll, jlgaddis, revertts, samcrawford, wahern, brian-armstrong, lrizzo, kev009, muppetman, Nrsolis, signa11, majke, shaklee3, emmericp, zamadatix, Sean-Der, techsupporter, downwithbgp, p1mrx, ghshephard, matsur

x/aws: colmmacc, _msw_, aligouri, illumin8, jcrites, socttlegrand2, openasocket, otterley, mslot, bbgm, NathanKP, appwiz, planckscnst, blasdel, mjb, ragona, donavanm, grogenaut, jeffbarr, twirrim, jtoberon, fnordpiglet, scarface74

x/nvidia: jebarker

x/google: kortilla

ai: jph00, eli_gottlieb, iandanforth, karpathy, mjn, nl, albertzeyer, bravura, michael_nielsen, dgacmu, cs702, emu, lhl, espadrine, edwardjhu, binarymax, jll29, MontyCarloHall

bootstrap: arvidkahl

crypto: jaekwon, tipsysquid, Taek, daeken, pbsd, davidcash

a11y: mwcampbell

eee: geerlingguy, sowbug, ta8645, mmmBacon, gchadwick, femto

Submissions

GELab-Zero: Android automation framework for multimodal LLMs

github.com
3 points·by ignoramous·há 17 dias·0 comments

Endo: JavaScript plugin framework with built-in supply chain attack resistance

github.com
2 points·by ignoramous·há 2 meses·0 comments

Toro: Deploy Applications as Unikernels

github.com
148 points·by ignoramous·há 6 meses·139 comments

Qwen3-VL 2B on Raspberry Pi with llama.cpp

eheidi.dev
2 points·by ignoramous·há 7 meses·0 comments

Apple and Foxconn in China

damagemag.com
1 points·by ignoramous·há 8 meses·0 comments

Toward provably private insights into AI use

research.google
2 points·by ignoramous·há 8 meses·1 comments

Xyne: Open-source LLM-driven search engine for Google Workspace

github.com
2 points·by ignoramous·há 9 meses·0 comments

comments

ignoramous
·há 9 horas·discuss
> the kind of engineering that goes into Xiaomi (and Deepseeks) inference optimizations

At Xiaomi, MiMo is now led by Luo Fuli. She is a former Alibaba & DeepSeek employee: https://newsen.pku.edu.cn/news_events/news/people/15385.html (https://archive.vn/I8Pmu) / https://e.vnexpress.net/news/tech/personalities/who-is-luo-f... (https://archive.vn/sb3B6)

Don't know if it is due to Luo, but it is striking how similar performance & pricing of the models, DeepSeek v4 Pro & MiMo v2.5 Pro, is.
ignoramous
·ontem·discuss
Curious why TFA calls out "Tencent in China".

  tencent/Hy3. New Apache 2.0 licensed model from Tencent in China
Is there a Tencent AI lab elsewhere (MiniMax have some association with Tencent, for example)?
ignoramous
·anteontem·discuss
[flagged]
ignoramous
·anteontem·discuss
Lower context window notwithstanding, Hy3's coding benchmarks hold their own against DeepSeek v4 Pro & MiMo v2.5 Pro. That's quite something for a model priced like DeepSeek v4 Flash & MiMo v2.5 (for non-cached tokens), which are 3x cheaper than their respective Pro variants.
ignoramous
·anteontem·discuss
Cheaper than Qwen 3.7 Max. Second indication, after Grok 4.5 ($2 in / $6 out), that the BigLabs are feeling the GLM 5.2 heat.
ignoramous
·anteontem·discuss
> This entire article is publicly and personally attacking him for choosing a different product.

Well, TFA has a conveniently titled section "Addressing the Blog Post", that raises (setting aside speculation) some good points:

  The [Bun rewrite] blog post is ... almost like the marketing department of a trillion dollar company has a lot of money riding on this article ...
  
  There's a dichotomy being presented here where you have to either choose a "style guide" or a programming language feature in order to avoid bugs. The sleight of hand misdirects the reader away from the main way bugs are eliminated: by dedicating engineering resources to it ...
  
  [TigerBeetle] put in the time to find and eliminate the bugs, they make an effort to maintain a healthy relationship with ZSF, and Bun did not do that.

  The argument for shipping all the million lines of unreviewed code is that the test suite is good enough to catch everything. Then why are you saying you have so many annoying bugs in the Zig code? What happened to the test suite being sufficient to catch everything ...

  Performance increase is attributed to LTO, which Zig has supported for all of Bun's existence. It used to be enabled by default until we ran into too many LLVM bugs, all of which also affect Rust ...

  The post claims they were fuzzing their Zig code, while during our calls the whole Bun team told us that they were not fuzzing anything. This appears to be an outright fabrication.

  The blog post outlines a bunch of engineering work done to reduce binary size, to better make the case that "Bun is better in Rust" ... you were doing the engineering work that you should have done in the Zig codebase since the beginning ...

  I noticed that you neglected to mention compilation speed. Zig compiler project is about 600,000 lines of code - roughly the same size as Bun before the rewrite, and I'm clocking 16s to build from scratch with a clean cache, followed by 90ms for each subsequent edit with incremental compilation enabled. What are the corresponding measurements of Bun post-rewrite?
ignoramous
·anteontem·discuss
> I suspect Claude might be faster and therefore cheaper, but maybe not by a lot.

While Jarred used Mythos-class model, some open weights, if they were as capable (certainly, GLM 5.2 looks the part), would have been way, way cheaper than professionals.

Approx costs:

  DeepSeek v4 Pro & Mimo v2.5 Pro   $3,426  ($2,567 / $600 / $259)
  Tencent HY3                       $3,892  ($1,180 / $552 / $2,160)
  GLM 5.2                          $30,016  ($8,260 / $3,036 / $18,720)
  Qwen 3.7 Max                     $37,925  ($14,750 / $5,175 / $18,000)
  Claude Opus 4.8 & GPT 5.5 xhigh  $82,750  ($29,500 / $17,250 / $36,000)

  5.9 billion uncached input tokens, 690 million output tokens, 72 billion cached input token reads.
ignoramous
·há 5 dias·discuss
What does it mean when prompting SoTA LLMs prone to slop to be concise and precise, with respect to context at hand, not work at all? Anyone benchmarking that?
ignoramous
·há 5 dias·discuss


  When Workers Cache is enabled, every cacheable request to your Worker hits Cloudflare's cache first. If there's a fresh cached response, Cloudflare returns it directly — your Worker doesn't run, and you don't pay CPU time for it. On a miss, your Worker runs, and if your response is cacheable, Cloudflare stores it for the next request. The next request from anywhere on Earth can be served straight from cache.
Incredible! This is why I shoehorn all my server side usecases on to the Workers Platform. Cloudflare, since 2020 when I first went all-in, has consistently shipped features that reduce bills significantly (except for 2023 Workers usage model changes). In one case, when they shipped free Snippets (Workers but 32kb code size & 5s CPU time) for Pro accounts ($200/yr), our bills went from £15k+ to £0.

I know about the infamous "Enterprise plan" (especially, when your bandwidth is as high as ours in 100s of TBs) and know of at least one other tech shop that was required to pay for it ... but we haven't got that sales call, yet.
ignoramous
·há 6 dias·discuss
> By definition you can't protect yourself from the entity that provides you the software you use, because you have now way to guarantee that they aren't going to backdoor you.

There are levels to this (hermetic/reproducible builds & attestations, as one example). In fact, TFA is harsh for the only case it wants to make an example out of, WhatsApp Web:

  Code Verify works in partnership with Cloudflare, a web infrastructure and security company, to provide independent, third-party, transparent verification of the code you're being served on WhatsApp Web. We hope this gives at-risk users peace of mind. 
https://blog.cloudflare.com/cloudflare-verifies-code-whatsap... & https://engineering.fb.com/2022/03/10/security/code-verify/
ignoramous
·há 9 dias·discuss
> much higher base of real intelligence

Not sure how much "real intelligence" is to be found in Mythos & Sol, but at this point, ignoring the intelligence gap, I find it totally impressive that the likes of GLM, Kimi, Qwen, MiMo hold their own at 2x to 4x less cost, and work for my use case just the same.
ignoramous
·há 9 dias·discuss
Interesting. If I may: What was this "clean up" pass? A code review? A code review with specialized prompt? A focused review to check for edge cases / logic errors / api misuse? Or, something else specific to the codebase?

Have you settled on what the clean up pass should look like? Or, do you keep experimenting with it?

In case one might not have been aware: Composer 2 was Kimi Base 2.5 post-trained (RL'd) by Cursor: https://news.ycombinator.com/item?id=48507474. Composer 2.5 might be something totally different.
ignoramous
·há 9 dias·discuss
Copilot Chat supports BYOK since Oct 2025 for the VSCode plugin: https://code.visualstudio.com/blogs/2026/06/18/byok-vscode
ignoramous
·há 9 dias·discuss
> Fable wrote it when it spawned the background task. How am I supposed to rephrase it?

Can the harness to auto-rephrase? I imagine, doing so will burn through tokens though.
ignoramous
·há 10 dias·discuss
Probably busy mirroring binary search trees? ;) https://news.ycombinator.com/item?id=9695102
ignoramous
·há 11 dias·discuss
> Google doesn't even have enough resources to decently deploy a model like that

Probably not for free but tbf, Google did scale "AI Mode" globally to its billion+ users, with its Gemini 3 series. Pretty much broke my habit of searching the web with pplx & Chat.
ignoramous
·há 11 dias·discuss
Amazing! Thanks for sharing.

> I didn't look at any of the code, but I had it build me a bunch of various visualization and tuning UIs. I was basically acting as a PM/TPM/QA engineer, and what I was able to do in a couple evenings is stuff that entire teams used to spend weeks on.

Agree. The pace of improvement (for coding tasks) from all Labs vying to catch up to the Big 3 (GDM, OAI, Ant) has been relentless.

> So I asked my agent (I'm using some free agents that are pretty decent - Nemotron Ultra from OpenRouter and Big Pickle from OpenCode Zen) to build me an OpenCV classifier to try to read the digits.

I think, Meta's Muse Spark is also free to use, for now. I'd be interesting if you'd ever find yourself using that.

If I may: Did you find Big Pickle is better than Nemotron Ultra or vice versa? Which harness (or IDE) did you drive these models with?
ignoramous
·há 11 dias·discuss
> Mullvad believes privacy is a universal right. You might disagree but at least it's consistent ...

Is privacy inconsistent with right to dignity, liberty, and equality?
ignoramous
·há 12 dias·discuss
> Educated people wouldn't go in populist direction.

Depends on "education".
ignoramous
·há 12 dias·discuss
Connection "reset": https://en.wikipedia.org/wiki/Transmission_Control_Protocol#...