Tropical engineer - Tropical blog

The Fable 5 is my new default for planning and brainstorming.

# Telegram IPv4+IPv6 address-list for MikroTik RouterOS # Updated: 2026-03-24 | Sources: BGP RIB + RIPE DB + cidr.txt /ip/firewall/address-list add list=telegram address=149.154.160.0/20 comment="TG DC1-5 AS62041/59930/62014" add list=telegram address=91.108.4.0/22 comment="TG AS62041 AMS" add list=telegram address=91.108.8.0/22 comment="TG AS62041 AMS" add list=telegram address=91.108.12.0/22 comment="TG AS59930 MIA" add list=telegram address=91.108.16.0/22 comment="TG AS62014 SIN" add list=telegram address=91.108.20.0/22 comment="TG AS44907 BOM" add list=telegram address=91.108.56.0/22 comment="TG AS62041/62014 SIN" add list=telegram address=95.161.64.0/20 comment="TG AS62041 infra" add list=telegram address=109.239.140.0/24 comment="TG legacy GNM/AS31500 AMS" add list=telegram address=5.28.192.0/18 comment="TG EU infra IPFFM/AS34665 FRA" add list=telegram address=185.76.151.0/24 comment="TG CDN AS211157 HEL" add list=telegram address=91.105.192.0/23 comment="TG CDN AS211157 HEL" /ipv6/firewall/address-list add list=telegram address=2001:67c:4e8::/48 comment="TG AS62041 AMS" add list=telegram address=2001:b28:f23d::/48 comment="TG AS59930 MIA" add list=telegram address=2001:b28:f23f::/48 comment="TG AS62014 SIN" add list=telegram address=2001:b28:f23c::/48 comment="TG AS44907 BOM" add list=telegram address=2a0a:f280:203::/48 comment="TG AS211157 HEL" add list=telegram address=2001:b28:f23c::/46 comment="TG aggregate MIA/SIN/BOM" add list=telegram address=2a0a:f280::/29 comment="TG RIPE alloc future-proof" 🛟🇷🇺

I used to think routing packets was complex. But routing meaning is much harder. The long silence taught me one big lesson: You cannot "patch" a wrong foundation with better commands. The old context is like a bad static route —> it stays in the cache and infects the new system. I discovered that the invisible Context is the ultimate operating system. To fix the core, do not just edit the config. Burn it down. The syntax is English now, but it compiles into pure logic. I am operating on a new layer now. Posting from: 🇻🇳 Vietnam, Da Nang city ☁️🌧️ 🌊 22° C

Make Vietnam great again! ✊ Fight! Fight! Fight! Posting from: 🇻🇳 Vietnam, Da Lat 🌧️🌧️ 19° C

Make Vietnam great again! ✊ Fight! Fight! Fight! Posting from: 🇻🇳 Vietnam, Da Lat 🌧️🌧️ 19° C

OpenAI GPT4 Turbo 🔥 🔥 🔥 128k tokens context window 🦾🤯 Knowledge cutoff April 2023 🫶 Ссылка скрыта

Dalle 3 rocks. This kind of detailed prompt was generated by ChatGPT4 to achieve the necessary vibe. Whenever I ask ChatGPT4 to help me with building prompts for AI, I always ask to generate detailed and comprehensive prompt to provide better context for AI. Prompts for Dalle 3: “This image offers a striking aerial view reminiscent of South Vietnam's famed kite beaches. The crystal-clear, aqua-blue waters of the ocean stretch far into the horizon, blending turquoise hues with deeper shades of blue. The pristine white sandy shore is adorned with rows of swaying palm trees, casting gentle shadows under the sun. Throughout the waters, vibrant kite-surfers are seen, their colorful kites dancing in the sky and adding a dynamic contrast against the backdrop of the serene sea. In the foreground, a wooden desk under a thatched roof provides a modern touch to the traditional landscape. Atop the desk, a laptop displays a digital interface, and it's flanked by an organized set of books, a green plant, and a mousepad. The scene perfectly captures the essence of South Vietnam's beaches, where the tranquility of nature meets contemporary life.” … Just my initial experiments with personal static website fully made by AI: structure, content, headers, graphics, HTML and CSS code. Let’s see how it works together 🙃 Posting from: 🇻🇳 Vietnam, Mui Ne ☀️ 31° C

Today, let's dive into another layer of ChatGPT. I'm sharing insights from a question that intrigued me during my recent experiments: How does ChatGPT-4 handle branching contexts? The Core Question While experimenting with ChatGPT-4, I found myself puzzled about its handling of edited or branching queries. Does the model recognize all versions, creating a sort of 'contextual tree'? This question is more than academic curiosity for me; as someone who runs a MikroTik training center, understanding complex technical behaviors is part of my daily life. Roots of the Query ChatGPT-4 has a certain limit of tokens (words, spaces, etc.) that it can process, often giving preference to the most recent inputs. This led me to question if this model recognizes branching paths or if it merely operates on the most recent iteration of the conversation. The Unveiled Truth My inquiries yielded fascinating insights. Each branch of the conversation knows only about its direct lineage, with no awareness of parallel conversations. This is unlike a tree that knows about all its branches; here, each branch is like its own entity. Implications for User Interaction Though your interface might allow you to flip back through conversation history, it's not affecting the model's understanding. Every branch operates as an independent context, aiming for the most accurate and specific response based on the most recent inputs. Final Thoughts Just like any technology, the more you delve into ChatGPT-4, the more intriguing complexities you find. This particular insight has allowed me to better understand how to manipulate queries for more accurate responses. While today we've merely scratched the surface, I'm eager to continue this exploration into the depths of AI's capabilities. #ChatGPT #ContextualAwareness #BranchingContext Posting from: 🇻🇳 Vietnam, Mui Ne 🌦️ 29° C

Today, we live in a digital world, and the magic of Artificial Intelligence (AI) is something that catches eyes of many. We know ChatGPT-4, but it's just a baby step. What if AI could hold a much bigger context, say near 1 million tokens in the next five years? Let’s dream a little. 1. Context Growth: It’s not just about more tokens. It's about making AI understand complex stories, making learning more tailored, fun, and reachable for everyone. 2. Changing the School Map: Today, where you live decides where you learn. But what if AI could change this old tale? I dream of a day when physical schools give way to digital classrooms, driven by smart AI, understanding each young mind's unique way of learning. 3. AI: The New Teacher: Imagine an AI that could offer a personal touch to each kid's learning journey. No matter where they are, they could reach for the stars with a world-class education tailored just for them. 4. My Steps Towards the Future: As a father and the founder of a networking academy, I am taking steps today for this beautiful tomorrow. By bringing AI tools into our teaching, we are not just talking networks; we are opening doors for young minds to a world where learning knows no limits. 5. Humans and AI: Learning Together: The bond between human teachers and AI could create a space where curiosity flies free, and knowledge is shared in a way that touches every learner, going beyond the old classroom walls. Final Thoughts: As I dive deeper into the world of AI and education, the dream of a big change keeps me going. It won’t be easy, but the idea of building a world where learning is for everyone, with AI’s helping hand, is a beautiful story waiting to be told. The seeds we plant today will help grow the education of tomorrow. #AIEducation #DreamingFuture #ContextualGrowth #LearningWithoutLimits Posting from: 🇻🇳 Vietnam, Mui Ne, Impresso co-working space 🌞☁️ 31° C