Dagens Vibes — 19. oktober 2026

Fra X-feedet

Satoshis bitcoin.exe genskabt byte for byte

T4T5 siger, at Opus 5.5 hjalp med at udlede det ukendte buildmiljø bag Bitcoin fra 2009, så den genskabte fil får samme SHA-256. Et skarpere succeskriterium end “ser rigtigt ud”; selve buildopskriften er ikke med i opslaget.

T4T5@t4t5

i rebuilt satoshi's 2009 bitcoin.exe byte for byte! afaik nobody has been able to perfectly recreate the executable published in 2009 from source, because we didn't know the exact build environment that was used. using opus 5.5, i reverse-engineered satoshi's build machine from the binary. it now outputs an exe with the exact same sha256 as the 2009 release!

♥ 1,304 · ↻ 75 · 💬 59 · 🔖 329

https://x.com/t4t5/status/2111898332658544835

DeepSeek Flash får konkrete browser-tests

Browser Use måler V4.1 Flash-1017 til samme 75/100 som Opus 5.5 for $0,22 mod $2,35 pr. opgave — cirka 10,7× billigere, ikke opslagets 19×. Justin Schroeders 3D-test var derimod langsom og ineffektiv: modellen skal vurderes i det arbejde, den faktisk skal udføre.

Browser Use@browser_use

DeepSeek V4.1 Flash 1017 just matched Opus 5.5 on long horizon browser use. 19x cheaper 👀 > 75/100, same as Opus 5.5, almost Astra level > $0.22/task vs $2.35/task > Sol it still SOTA, but it's close The new pareto frontier 🔥

♥ 1,260 · ↻ 83 · 💬 63 · 🔖 337

https://x.com/browser_use/status/2111696102106038319

Justin Schroeder@jpschroeder

DeepSeek V4 Flash 1017 is certainly better at 3d than previous DeepSeek models. It was also one of the slowest and least efficient models yet.

♥ 0 · ↻ 0 · 💬 1 · 🔖 3

https://x.com/jpschroeder/status/2111932155873869992

Tre dages specifikation, én dags implementering

Banteg fik netplay ind i et spil, der ikke var bygget til det: først 3–4 dage med arkitektur og modeldiskussioner, så Opus med Sol/Argon-review. Hans tidligere forsøg strandede i debugging; forskellen var arbejdet før koden.

banteg@banteg

i just shipped netplay in a game whose architecture was never planned to accommodate it. my time was split as follows: 3-4 days of grilling frontier models on the problem space and drafting the spec. over this time i found good answers to every single problem and my skepticism gradually dissolved. i went from "it's impossible for it to play well with 500 active entities to sync" to "it may be actually feasible and even rather cheap and it would play well even across continents". it's a bespoke design that fits this game specifically, it reuses the strong parts about my previous rearchitecture and it's stripped to the core, reusing as much of the existing game machinery. 1 full day of opus implementing the spec, along with multiple rounds of reviews by other models (sol, argon). i used ultra effort and everything worked flawlessly when it was done. this differs quite a lot from multiple previous attempts to add netplay where things went wrong and i got stuck with nasty debugging. this time i put much more human effort (asking the right questions) upfront and got an amazing result.

♥ 155 · ↻ 3 · 💬 15 · 🔖 34

https://x.com/banteg/status/2111844254351831406

Et testmiljø agenten selv kan bruge

Adams demo lader agenter betjene appen, se og høre resultatet og forbedre både produkt og testmiljø. Det interessante er feedbacksløjfen, ikke endnu et værktøj, der kan klikke på en knap.

Adam@_overment

the environment for self-play is now the critical part of software. below is my progress on building one for a project. it lets agents interact with the app, see, listen and verify their work. they can also observe the system and self-reflect to improve not only the app but also the environment itself.

♥ 51 · ↻ 0 · 💬 5 · 🔖 63

https://x.com/_overment/status/2111741320528109794

Videoredigering med Blender og sky-rendering

Riley Brown beskriver et Opus-flow fra rå optagelse til B-roll, klipning, Blender-scener og færdig video. De knap 20 minutter og under $1 gælder GPU-renderingen — ikke hele arbejdet; sceneskabelonen var allerede på plads.

Riley Brown@rileybrown

This video was edited with 1 prompt with Opus 5.5. I uploaded a single video of myself (with many mistakes). Opus went on to: 1. Analyze the transcript 2. Search Twitter for the relevant B-roll (Claude Code literally calls GrokBot for this step) 3. Scrape the Twitter video assets 4. Cut the video based on mistakes and word gaps 5. Build out the Blender scene from a template 6. Build the UI and components of each shot (my Blender scene has 4 shots: a computer, phone, table, and board) 7. It took screenshots of every section, analyzed and fixed them 8. Synced every shot, zoom, and animation to the exact words I say 9. Animated a robot arm placing real logos on the table right when I mention them 10. Drew the whiteboard ending with logos and graphics instead of text 11. Pulled a cutaway from the creator's own video, with his voice 12. Rendered all 3,487 frames on RunPod cloud GPUs in under 20 minutes, for less than $1 13. Put me in the bottom third only when I'm talking, so the B-roll stays full screen 14. Added word-by-word captions, sound effects, and music 15. Uploaded it to my Google Drive, ready to post Will be making a full youtube video this week explaining it. I want to learn a little more about how cloud rendering works before making video, as well as figuring out what the best provider is. (If you are in cloud rendering space, hit me up) My goal of the youtube video will not be to make it a copyable skill, rather teach people the foundations so they can be creative and make their own mini-world by which they can make videos from, while understanding what's happening under the hood. Shout out @_chuckyc @_MaxBlade

♥ 227 · ↻ 8 · 💬 51 · 🔖 303

https://x.com/rileybrown/status/2111925755215040779

Nyhedsbonus

Disney tilbyder job til størstedelen af Character.AI

Efter konkursbegæringen tilbyder Disney job til mindst 60 ansatte og vil købe aktiver for $20 mio., skriver TechCrunch ud fra retsdokumenter. Det er et foreløbigt bud, som andre kan overbyde — en talenthandel, ikke en redning af chatbotforretningen.