Posts

Introduction A MacBook Air M4 has 16GB of unified memory. Train a 3B model with PyTorch and the fan spins up within minutes; on the fanless model, thermal throttling kicks in. Bit-Axon is a 3.2B parameter hybrid language model that solves this constraint at the architecture level. The core idea is a three-layer sandwich structure: 24 layers divided into three segments, each using a different computation paradigm. L L L a a a y y y e e e r r r 1 1 9 7 - - - 8 1 2 : 6 4 : : █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ S S S P W M u A r + e + M A M o x o E o E n - S S M → → → C D O o e u n e t t p p e u x r t t e a s a s y b o n s n t o i h r n e p g s t i i ( s o O n ( ( n l ( ) i O n ( a e 1 t a ) t r e m n + e t m i s o o p r n a y ) r ) s e ) This isn’t just an intuitive division. Each segment addresses one of the three fundamental limitations of the Transformer architecture — quadratic complexity, memory explosion, and compute density. This post covers the mathematical foundations of each layer group, MLX framework optimizations, and thermal-aware training — the complete design for running an LLM on a MacBook. ...

Introduction Developing cloud-native applications comes with recurring pain points: AWS calls in CI pipelines incur costs, development stops without VPN, and onboarding requires credential setup. DevCloud is a local AWS emulator that solves these problems entirely offline. It achieves 671/699 test cases passing (96%) in boto3 compatibility tests. That number isn’t just a test pass rate — it’s a measure of how precisely the protocol layer replicates AWS behavior. This post explains how a single Go binary achieves this level of compatibility, covering protocol detection, serialization challenges, the plugin architecture, and the path from 96% to production-ready. ...

Introduction AWS provides over 200 services, each with its own API protocol and request/response structure. Manually implementing them one by one is practically impossible. At DevCloud, we built a codegen pipeline that reverse-engineers AWS’s internal modeling language, Smithy, to auto-generate Go code for nearly all AWS services. This post covers the full flow from parsing Smithy models to generating Go code, and how the auto-generated code powers a local AWS emulator. But the more fundamental question is: why generate code from an IDL at all? To understand the value beyond simple “productivity automation,” we need to first understand the problem this approach solves. ...

Introduction KVS v1.0.0 has been released. KVS is a simple in-memory key-value store written in Go that can be used as a Go module or deployed as a standalone server. This post introduces the major features included in v1.0.0 and takes a deep dive into the core data structures: Red-Black Tree and LSM Tree implementations. Why Another Key-Value Store? Excellent key-value stores like Redis, LevelDB, and BoltDB already exist. So why build KVS? KVS started as a learning and experimentation project. The goal was to experience firsthand the design decisions and trade-offs involved in building a production-grade database. The result is a store with these characteristics: ...

Introduction ACOR is a Go library that implements the Aho-Corasick algorithm with Redis as the backend storage. The latest version introduces four major features: Index APIs - Provides match position information Redis Topology Support - Supports Sentinel, Cluster, and Ring Command-Line Tool - Use directly from the terminal Server Adapters - Deploy as HTTP and gRPC services This post covers the usage and features of each. Index APIs Comparison with Existing APIs Previously, Find and Suggest APIs only told you which keywords matched. For text highlighting or position-based analysis, you had to calculate indices separately. ...

Introduction to Ansible Molecule with KIND Learn how to set up a Kubernetes automation testing environment by combining Ansible Molecule with KIND (Kubernetes IN Docker). Overview Ansible Molecule Ansible Molecule is a framework that helps test Ansible Roles in isolated environments using virtualization technologies. It supports various drivers and can integrate with KIND using the Delegated driver for Kubernetes environments. KIND (Kubernetes IN Docker) KIND is a tool that runs Kubernetes clusters as Docker containers. It allows you to quickly and easily create Kubernetes clusters locally, making it useful for: ...

Introduction to Molecule Molecule is a testing framework for Ansible Roles maintained by the ansible-community. With Molecule, you can systematically test Ansible Roles, enabling comprehensive testing using multiple instances, operating systems, virtualization providers, test frameworks, and test scenarios. Why Do You Need Molecule? When developing Ansible Roles, you face the following challenges: Limitations of Manual Testing: Manually running and verifying Roles each time is time-consuming. Multi-Environment Support: You need to ensure Roles work correctly on various operating systems like Ubuntu, CentOS, and Debian. Continuous Integration: Tests need to run automatically in CI/CD pipelines. Code Quality: Ansible code quality must be maintained consistently. Molecule provides the following features to address these challenges: ...

I’m documenting terms I’ve encountered or learned while communicating and working in the workplace, so I won’t forget them and can refer back to them occasionally. This collection covers terminology used not only in software engineering but across business in general. Table of Contents MECE Dogfooding ISO 8601 Ice Breaking Housekeeping Job On-the-fly SLA/SLO/SLI Dogpile Effect Thundering Herd Zero Trust Shift Left Technical Debt MECE Source: Logical Analysis MECE without Duplication and Omission ...

Introduction I’ve released ACOR v0.2.0. This version focuses on restructuring the project to follow Go standards and fixing a few bugs. Standard Project Structure v0.2.0 restructures the project to follow the standard Go project layout (#2). As the project grew, managing the file structure systematically became necessary. Following the Standard Go Project Layout guidelines: pkg/: Package code importable by external projects internal/: Private application code cmd/: Main applications This structure makes the project’s intent clearer and more maintainable. ...

Introduction I’ve released ACOR v0.1.0. This version focuses on modernizing the project foundation rather than adding new features. Following the evolution of the Go ecosystem, I migrated the dependency management and CI/CD systems to current standards. Why Go Modules ACOR initially used Glide for dependency management. Glide was widely used in the community during the era when Go lacked an official dependency management tool. However, things changed when Go modules were officially introduced in Go 1.11. ...