Vlad Shurupov

Passionate about #Philosophy, #ArtificialIntelligence, #AlgorithmicTrading, #TradingStrategies, #Coding and lots of other exiting things.

[Vlad Shurupov @AlgoDesigner]

@EleonoraSayaka Both!

[Vlad Shurupov @AlgoDesigner]

I came across an engaging interview by Eleonora Sayaka with Kazuhiko Nishi, the quiet visionary behind the MSX. The very MSX that greatly influenced my choices in life. It let me experience magic early on and apply it in new areas. Nishi's journey is both educational and thought provoking. It is a lesson in bravery, passion and dedication.

https://t.co/6yVMmEimdH

#MSX #Computers #Programming #RetroComputers #MSX3

@EleonoraSayaka
@nishikazuhiko

[Vlad Shurupov @AlgoDesigner]

@EleonoraSayaka I have a brand new SVI-328 in my retro collection. It is a nice piece of MSX history.

[Vlad Shurupov @AlgoDesigner]

@Saas_addy Panasonic FS-A1GT MSX turbo R
#MSX

[Vlad Shurupov @AlgoDesigner]

@nishikazuhiko このパソコンは間違いなく買い物リストに追加します。

[Vlad Shurupov @AlgoDesigner]

@karthikponna19 Forth

[Vlad Shurupov @AlgoDesigner]

@GameStalgiaX We did and still do. Just take it out of the closet and fire it up!

[Vlad Shurupov @AlgoDesigner]

AI coding agents are powerful but fundamentally blind to your codebase structure. When your agent edits validate_token(), it has no idea that 47 callers depend on its return type. When it searches for "database connection", it greps blindly through every file. Without a code graph, your agent works like a surgeon operating without an X-ray, skilled but guessing at what's inside.

codeloom builds a queryable code graph from your entire codebase, every function, class, import, call, and document, and exposes it to your AI agent. One install, and your agent stops grepping and starts understanding.

https://t.co/aYGymIuHTL

#AIAgents #DevTools #CodeGraph #Coding #LLM

[Vlad Shurupov @AlgoDesigner]

"𝑨𝒕 𝒕𝒉𝒆 𝒄𝒆𝒏𝒕𝒓𝒆 𝒐𝒇 𝒚𝒐𝒖𝒓 𝒃𝒆𝒊𝒏𝒈 𝒚𝒐𝒖 𝒉𝒂𝒗𝒆 𝒕𝒉𝒆 𝒂𝒏𝒔𝒘𝒆𝒓; 𝒚𝒐𝒖 𝒌𝒏𝒐𝒘 𝒘𝒉𝒐 𝒚𝒐𝒖 𝒂𝒓𝒆 𝒂𝒏𝒅 𝒚𝒐𝒖 𝒌𝒏𝒐𝒘 𝒘𝒉𝒂𝒕 𝒚𝒐𝒖 𝒘𝒂𝒏𝒕." - 𝑳𝒂𝒐 𝑻𝒛𝒖
Most of our problems are fundamentally the same. They simply require you to be in Zen mode: to keep things simple and solve them patiently, one by one. #Philosophy #Life

[Vlad Shurupov @AlgoDesigner]

@nevets1352 @alphafox Are there YouTube videos on the construction process? This is an engineering materpiece.

[Vlad Shurupov @AlgoDesigner]

@Shivam25mishra What is Visual Studio?

[Vlad Shurupov @AlgoDesigner]

@heyyyyyieeee C

[Vlad Shurupov @AlgoDesigner]

Using rapid agent-based development in complex, mission-critical systems requires laser-like focus, attention to seemingly innocent details, and an almost supernatural amount of patience.

You have to understand what LLMs are and what they are not. That means ignoring the hype and diving into the gritty implementation details, hardships and all.

If you're tempted to never return to your IDE or text editor, enjoying the presumed comfort of the architectural throne... you're in for a big surprise. Your house of cards will crumble under the weight of technical debt.

Scary, isn't it? The lengths to which machines will go to please an ignorant master.

#AgenticAI #LLMs #SoftwareArchitecture #TechnicalDebt #MissionCriticalSystems

[Vlad Shurupov @AlgoDesigner]

Agents are improving fast, but how they acquire knowledge remains prehistoric: glob and grep. The less they know, the more you pay in tokens.

I've been hunting for a tool that could build a solid graph covering our company's entire knowledge base. Everything I tried hit the familiar walls: over-engineered complexity, vendor lock-in, no local-first option. Privacy was non-negotiable.

Forked the project closest to the vision and dived into coding: hybrid search, incremental builds, better relevance scores, source-vs-test prioritisation. It already cuts through a lot of noise grep subjects you too.

The pencilled roadmap is Agents → Knowledge Base → Advanced memory → Dream Machines…

#RAG #KnowledgeGraph #LocalFirst #AgenticMemory #TokenEfficiency #AI

[Vlad Shurupov @AlgoDesigner]

@Visual_Salman I have seen it and I love it! 🙂

[Vlad Shurupov @AlgoDesigner]

This video finally came off my to-watch-and-analyse list. In the darkness of the autumn evening, I watched.

The similarities between teaching a machine to play Atari games and, in our case, "higher frequency" algorithms are striking.

(1) The "Reality" of Non-Turn-Based Environments.
Carmack stipulated that reality is not a turn-based game. In standard RL the environment waits for the agent, but in the "inverted RL environment" Carmack proposes, the environment keeps going regardless of whether the agent has finished computing. This directly mirrors financial markets, where the order book and price action move continuously. Your ANN may be slow (hence my preference for shallow networks). If an algorithm takes too long to decide, the opportunity has already changed.

(2) The Critical Impact of Latency.
Carmack described latency as a killer for many algorithms. He notes that state-of-the-art algorithms condition their world models on the immediate next action and observation, which works in a simulator but falls apart when there is an action delay of even a few frames. In trading, end-to-end latency from sensing the market to executing the action (new order single, amend, cancellation) determines whether the agent's reward signal is meaningful or just random noise.

Cross disciplinary research rocks. More later.

https://t.co/HRLApNjMGe

#ReinforcementLearning #AlgorithmicTrading #Latency #JohnCarmack #HighFrequencyTrading #Trading

@ID_AA_Carmack

[Vlad Shurupov @AlgoDesigner]

@ravikiran_dev7 Ignore the white noise and stay on target. And "may the Force be with you!" 🙂

[Vlad Shurupov @AlgoDesigner]

@hck_lab I admire your passion for AI. If you would like to connect, look me up on LinkedIn.

[Vlad Shurupov @AlgoDesigner]

Mount a FAT12 MSX HDD image as a real filesystem,
edit files on your host,
and watch them appear instantly in openMSX.

No disk juggling. No manual steps. Just flow.

👉 https://t.co/dgctTn6jkB

One image. Two worlds.

#MSX #RetroComputing #Emulation

[Vlad Shurupov @AlgoDesigner]

@Palak3312 Windows.