· ᴀᴅᴀᴍ ʟᴀɴɢᴇ ·
29.4K posts

· ᴀᴅᴀᴍ ʟᴀɴɢᴇ ·
@AdamLangePL
Breaking things professionally | Researcher | Speaker | Lecturer | Podcaster | #ProudToBeDefender | Opinions are my own.
0x90 Katılım Mayıs 2012
588 Takip Edilen4.4K Takipçiler
Sabitlenmiş Tweet

@AdamLangePL Cena czyni cuda (za apple tv), może nawet za 800-1000 zł. Tania maszynka do przeglądania internetu. :)
Polski

"Solid" ;) it's just mumbo-jumbo mix of wishes. No serious leak will say "Faster GPU" or "Improved Gaming" or "Better HDR" ;)
get lost!
Apple Club@applesclubs
📺 Apple TV 4K (2026) leaks look solid! • A17 Pro chip 🔥 • Ray tracing + faster GPU 🎮 • Apple Intelligence + smarter Siri 🧠 • Wi-Fi 7 & Bluetooth 6 ⚡️ • Better HDR, same design And possibly under $100… this could be Apple’s sleeper hit 👀
English

@AdamLangePL Gdyby tak dało się przerobić na ustrojstwo do przeglądania internetu w przeglądarce z obsługą klawiatury i myszki. Mini mac mini :)
Polski

@boczemuniepl Nie ma co się podniecać. To nie "leak" tylko kompilacja życzeń zbudowana na bazie poprzednich kompilacji życzeń ;)
Polski

Jeśli to się potwierdzi to dosłownie Apple zaora kolejny (i tak już słabiutki) rynek dostawek TV. Zwłaszcza, że 90% znajomych pomimo smart TV własnego telewizora i tak dziś już woli Apple TV z uwagi na tvOS, który po prostu działa.
Apple Club@applesclubs
📺 Apple TV 4K (2026) leaks look solid! • A17 Pro chip 🔥 • Ray tracing + faster GPU 🎮 • Apple Intelligence + smarter Siri 🧠 • Wi-Fi 7 & Bluetooth 6 ⚡️ • Better HDR, same design And possibly under $100… this could be Apple’s sleeper hit 👀
Polski

@imag0r Yup, i noticed. For me its just experiment ;) i promise i will not publish it!
English

First test of "Creative writing" by GPT-OSS-120B on DGX Spark.
1:48 to create a book with 8 chapters and about 30k words in Polish.
Pretty good result (readable, consistent) but with a bit of glitches here and there.
Changing code to add "reviewer" actor that does more complicated review on final book and fix if needed.
Next in line - use Polish grammar/style model to refine sentences.
This is pretty interesting experiment :)

English

GGUF FP16 version of Allegro/PLT5 model that i build for own tests. Maybe it will be useful for some :)
huggingface.co/AdamLangePL/al…
English

Yes. AGI has been achieved. We are now working on its levels to reach superintelligence (ASI).
The Verge@verge
Nvidia CEO Jensen Huang says ‘I think we’ve achieved AGI’ theverge.com/ai-artificial-…
English

· ᴀᴅᴀᴍ ʟᴀɴɢᴇ · retweetledi

She looks for what others assume can’t be found - Agata Ślusarek joins the #CONFidenceConf 2026 lineup!
• OSINT & threat analysis
• Telegram investigations
• Cybercrime + AI
Don't miss her lecture - join us:
🎟eventory.cc/event/confiden…

English

@maciej_szymczyk Yes, less quality than gpt-oss, and also loops frequently
English

I was thinking about having local llm (on dgx spark) or going with API (lets say Anthropic). For my estimate usage it looks like api cost (monthly) would be 1.5-3k usd (yeah…). So it looks like having local dgx is … cheaper with more or less good quality models.
Coding: GPT-OSS-120B
Vision + simple tasks: qwen3-vl-30b
GIF
English

@maciej_szymczyk It tends to loop with agentic tasks (coding), and oss-120 gives better results overall :) but qwen is good for summarizing, data extraction and visual recognition
English

@AdamLangePL I'm experimenting with qwen3.5 27b and results are stunning
English

· ᴀᴅᴀᴍ ʟᴀɴɢᴇ · retweetledi







