Søren Reinke

7.3K posts

Søren Reinke banner
Søren Reinke

Søren Reinke

@theDiverDK

IT Instructor, Microsoft Certified Trainer, karate instructor, work with AI/LLM. And proud dad of 2 wonderfull kids ;)

Hillerød, Denmark Katılım Mayıs 2011
609 Takip Edilen317 Takipçiler
Søren Reinke
Søren Reinke@theDiverDK·
@CommodoreBlog There was only 1 good joystick for the Amiga and C64, the Competition Pro 5000
English
0
0
1
33
CommodoreBlog
CommodoreBlog@CommodoreBlog·
New email from #Commodore this morning with a hint on a new joystick coming? Quickshot 2 Turbo please!🕹
CommodoreBlog tweet media
English
17
9
111
4.4K
Søren Reinke
Søren Reinke@theDiverDK·
@CNC_Kitchen Any chance you could update your excellent bumpmesh tool, so that it could create a 'see through mesh', like voronoi? Like the image i found?
Søren Reinke tweet media
English
0
0
0
21
Søren Reinke
Søren Reinke@theDiverDK·
@EgorBo @davidfowl And experience with the model you use. I have seen a lot of people using exact the same prompt for multiple LLM's, and getting quite different results. A prompts need to be optimized for the specific LLM (which I am very sure you know ;-) )
English
0
0
0
53
EgorBo
EgorBo@EgorBo·
@theDiverDK @davidfowl I had another benchmark today, opus did it for 45$, gpt for 220$. So it's hard to actually state anything. I guess it depends on the harness, instructions and pure luck :\
English
1
0
2
137
EgorBo
EgorBo@EgorBo·
Tired of toy AI benchmarks like making simple games with threejs. I had 3 LLMs (Opus 5, Opus 4.8, GPT 5.6 Sol) implement a real register allocator for .NET's Tier0 JIT instead. Goal: 2%+ faster compilation (measured with ~0.01% precision). The task involved some creativity and trade-offs. Opus 5 killed it.
EgorBo tweet media
English
4
7
55
6.1K
Søren Reinke retweetledi
Chayanka_42
Chayanka_42@42_gravity·
GPT-5.6 Sol Ultra helped solve a six-year-old open problem in quantum cryptography by generating the construction and main proof ideas through Codex. Human experts refined and verified the result.
Chayanka_42 tweet media
Henry Yuen@henryquantum

I'm happy to see the unclonable encryption problem solved; Prabhanjan Ananth and Amit Sahai report their findings, with the help of the UCLA harness (link below). This was a challenge for the quantum crypto community for the last 6 years.

English
15
102
1.1K
314.6K
GREG ISENBERG
GREG ISENBERG@gregisenberg·
Are you an Opus 5 person or a Fable 5 person?
English
314
4
356
128.4K
Søren Reinke
Søren Reinke@theDiverDK·
@unity Its been some time since I have used Unity, has it finaly build in support for GIT?
English
0
0
0
477
Unity
Unity@unity·
Introducing Unity 7 - built for a faster-moving, more open industry where creators, cross-functional teams, and coding agents can seamlessly collaborate across every stage of the development cycle. Featuring: - Create faster: Modernized CoreCLR foundation with a near-instant Play Mode and up to 90% faster shader builds. - Collaborate in better ways: New CLI, public API, and free MCP connect teams and coding agents directly to Unity. - Ship sharper graphics: Richer, realistic lighting via Surface Cache GI across all Unity supported devices. - Grow and monetize smarter: Built-in Unity Vector tools, native webshops, and direct-to-consumer IAP to scale efficiently. - No breaking changes: A direct continuation of Unity 6 architecture, offering an upgrade path with zero rebuilding. Unity 7 will enter early Beta testing in December, with the Full Release scheduled for Q1 2027.
English
61
404
2.1K
613.9K
Mike
Mike@mikechaves_io·
Just finished the OpenAI Dev Newsletter on the GPT-5.6 Build Kit. Here are the 6 things that actually matter if you’re building serious coding agents or agentic products: 1. Retire the giant master prompt OpenAI’s own data: leaner system prompts improved scores 10-15% while cutting tokens 41-66%. Move to layered AGENTS.md (root → nested) + Skills + task-only prompts. State each rule once. 2. Use /goal for anything with a measurable finish line Migrations, large refactors, experiments, games, retry loops. Give the agent a clear stopping condition and let it keep going until it’s proven done. 3. Subagents exist to isolate noise, not just multiply output Park the logs, stack traces, exploratory notes, and intermediate results in subagents so the main thread stays clean and focused. 4. Browser verification is now part of the definition of done If it has a UI (or generates a game), Codex should open it, interact, screenshot, and verify. Static code review alone is no longer enough. 5. Treat context as an explicit architecture - Cache the large, stable prefixes - Persist reasoning only while the objective is stable - Use Programmatic Tool Calling strictly for deterministic preprocessing Everything else stays in the main semantic loop. 6. Sites and Workspace Agents are support infrastructure Great for demos, experiment viewers, scheduled reports, and evidence boards. They are not your production runtime. The real upgrade isn’t “use fewer tokens.” It’s redesigning how instructions are layered, how long-running work is bounded, and how verification happens.
English
2
0
2
116
Mike
Mike@mikechaves_io·
Going to drop some really nerdy Codex + GPT-5.6 stuff in this thread. Learned a few savvy things from @OpenAI’s latest Dev Newsletter. Most of you already do this, but I’m a derp. Hoping it helps someone. 100% using AI going forward in this thread, so here it goes: 🧵
English
1
0
3
193
Søren Reinke
Søren Reinke@theDiverDK·
@Sonos Thanks for the update yesterday. before my system worked, now it plays for like 1-3 minutes and it stop when streaming from apple music on iPhone. Fix the SOFTWARE #bug #fixit #badsoftware
English
1
0
0
139
Søren Reinke
Søren Reinke@theDiverDK·
@steipete Hi Peter, I get when upgrading CodexBar: Error: Cask reports different checksum: 624172121dd792bf63e6074795e339dec83acf07792d463c28ca612f76c85217 SHA-256 checksum of downloaded file: 02d884074eed1f973fcb53d0fe82906e4eb25d9b96ac33cd07be7da540ebf81f
English
0
0
0
29
Søren Reinke
Søren Reinke@theDiverDK·
@berlingske Jeg undrer mig over medierne igen og igen gentager hvad den gamle mand siger. For det er aldrig noget fornuftigt. Han er den direkte årsag til de problemer vi har, ikke mindst med el nettet.
Dansk
0
0
0
21
Berlingske
Berlingske@berlingske·
En for hård retorik over for muslimer risikerer at skade integrationen, lyder det fra Socialdemokratiets tidligere formand Mogens Lykketoft. Og han mener ikke, at løsningen på S' vælgerflugt findes i at forsøge at efterligne Dansk Folkeparti. berlingske.dk/politik/lykket…
Dansk
238
11
51
38.7K
Mark Seemann
Mark Seemann@ploeh·
My 14th consecutive Microsoft MVP award: How lovely 😊
English
2
1
46
2.8K
Søren Reinke
Søren Reinke@theDiverDK·
@PerClausen3 Hvordan kan det være du konstant udtaler dig om ting hverken du eller dit parti har forstand på?
Dansk
0
0
0
18
Per Clausen
Per Clausen@PerClausen3·
Bruger mandag aften til at tale om, hvordan den grønne omstilling kan bruges til at skabe gode jobs i industrien og forbedre vores konkurrenceevne. Det gør jeg med udgangspunkt i erfaringer fra Aalborg. Desværre er flertallet i EU ved at opgive den grønne omstilling til skade for klima, miljø og natur, men også til skade for vores konkurrenceevne og muligheden for at skabe gode jobs i industrien.
Per Clausen tweet media
Dansk
15
1
8
812
Søren Reinke
Søren Reinke@theDiverDK·
@CasparStefani Nævn bare en positiv ting enhedslisten har bidraget med til samfundet, bare en enkelt……. nej jeg kan heller ikke finde en
Dansk
0
0
0
32
Caspar Stefani
Caspar Stefani@CasparStefani·
Se hvordan det “grønne byrum” bare bugner med mennesker her midt på dagen. Godt vi fik lukket den hovedfærdselsåre midt i byen. Man kan næsten ikke komme til for alle de mennesker. Det siger meget at selv træerne allerede er døde. Tak Enhedslisten ❤️
Caspar Stefani tweet media
Dansk
56
24
807
47.3K
Søren Reinke
Søren Reinke@theDiverDK·
@Nielsen88134K @k_torning Kan du komme med eksempler på han faktisk har passet sit arbejde? F.eks. sørget for at elnettet er fulgt med udviklingen…. nå nej det var jo et eksempel på han ikke har passet sit job!
Dansk
0
0
3
11
Keld Nielsen
Keld Nielsen@Nielsen88134K·
@k_torning Kom ned konkrete beviser på at Lykketoft ikke har gjort sit arbejde og ikke blot udfald grundet dit had til socialister som du hader så meget.Det lader dine kommentarer på X ingen tvivl om
Dansk
4
0
2
215
Kristian Tørning
Kristian Tørning@k_torning·
"Tidligere på året meldte Mogens Lykketoft ud, at han ikke genopstiller som formand for bestyrelsen." Lykketoft efterlader sig et syndigt rod. Vores forsyningssikkerhed er i fare. Og han efterlader et Energinet, der fejler kontinuerligt. Alligevel slipper Lykketoft for kritiske spørgsmål i pressen. kefm.dk/aktuelt/nyhede…
Dansk
14
7
114
6.5K
Sonia Sharma
Sonia Sharma@sonialy0·
Be honest, Ubuntu or Windows?
Sonia Sharma tweet mediaSonia Sharma tweet media
Indonesia
46
1
47
1.6K
Søren Reinke
Søren Reinke@theDiverDK·
@berlingske Er der evidens for noget som helst SF foreslår? De har ingen kontakt med virkeligheden
Dansk
0
0
0
28
Berlingske
Berlingske@berlingske·
SF vil bruge milliarder på velfærd uden at finansiere det med klassiske reformer, fortæller partiets finansordfører. I stedet skal investeringerne styrke økonomien. Men ifølge økonomiprofessor er der ingen evidens for, at det kan lade sig gøre. berlingske.dk/politik/sf-vil…
Dansk
22
4
32
3.8K