AI
OTP Bank Launches Self-Service AI Agent Framework
OTP Bank and red_mad_robot have launched a new architecture enabling business units to independently build and deploy AI agents.
October 15, 2026, 02:36 PM·CNews.ru
#software-architecture
AI
COSMOS: A New Organizational Layer for AI Agents
Nikita Krasnov has launched COSMOS, an organizational platform for AI agents designed to automate routine business processes and ensure verifiable results.
Хабр
Money
Giper.Dev Unveils Giper Baza Decentralized Database Infrastructure
The Giper.Dev guild has unveiled Giper Baza, a decentralized database architecture designed for offline-first operations and secure, zero-trust microservice communication.
Хабр
Hardware
OpenRTL: Replicating RAD Studio Architecture in C++
A developer has built OpenRTL, an open-source C++17 implementation of the RAD Studio VCL architecture, in just 1,500 lines of code.
Хабр
AI
Standardizing LLM Tool Definitions
A new proposal for a universal tool interface aims to eliminate the need to rewrite LLM tool wrappers for different frameworks.
Хабр
AI
Managing AI-Human Task Handoffs with Finite State Machines
Agent Inbox uses a finite state machine model and atomic database operations to manage human-AI task handoffs and prevent race conditions in multi-process environments.
Хабр
AI
Manus 2.0 Debuts Cascade Architecture to Improve AI Efficiency
Manus 2.0 introduces the Cascade architecture, reducing operational costs by 32% and enhancing task efficiency for the Singapore-based AI agent platform.
Crypto Briefing
Hardware
SilentJSON: A Streaming Approach to High-Performance Data Processing
The SilentJSON library for Go implements a streaming data processing model that prioritizes low latency and high throughput by avoiding full object decoding.
Хабр
AI
Structuring Project Semantics for AI Coding Agents
A lightweight Spec-Driven Development approach using four Markdown files helps AI coding agents understand project intent and reduce regressions in small repositories.
Хабр
Hardware
Solving the Domain Model Dilemma with Articulate ORM
The Articulate ORM library introduces a method to map multiple entity representations to a single database table, aiming to resolve the conflict between anemic and bloated domain models.
Хабр
AI
Why Model-Level Read-Only Restrictions Fail to Secure AI Systems
A laboratory test reveals that restricting an AI model's write access does not prevent the entire system from modifying CRM data if downstream components retain execution authority.
Хабр
AI
VECTA: Managing AI Agent Workflows and Git Integration
VECTA, a workspace for AI coding agents in closed testing, manages task execution and project state using Git-based snapshots and multi-level access control.
Хабр
AI
A Practical Guide to C4 Architecture Modeling
The C4 model offers a structured framework for software architecture visualization, utilizing four levels of detail to help teams document and communicate system designs effectively.
Хабр
AI
Veynrel: The Evolution of Vault Audit AI
The Obsidian plugin Vault Audit AI has rebranded to Veynrel, introducing a standalone Node.js companion service and an MCP-based architecture for secure AI interaction with local vaults.
Хабр
AI
Decoupling AI Agent Personality from Execution with PES
Researcher Yisen Xi has proposed the Persona–Execution Separation (PES) architecture to allow frequent updates to AI agent personalities without compromising strict audit requirements.
Хабр
Hardware
Strategic Approaches to Kotlin Multiplatform Adoption
Companies are shifting toward selective KMP implementation to optimize mobile development costs and reduce code duplication.
Хабр
AI
Inside the Architecture of AI Agents
AI agents like OpenClaw and Hermes use a cyclic architecture to connect LLMs with the real world through tool calls and memory layers.
Хабр
That is everything