- Programmer Weekly
- Posts
- Programmer Weekly (Issue 315 August 27 2026)
Programmer Weekly (Issue 315 August 27 2026)
Rahul Chaudhary
August 27, 2026
In partnership with
Welcome to issue 315 of Programmer Weekly. Let's get straight to the links this week.
100+ Claude Code hacks to ship code 10X faster
Top engineers at Anthropic and OpenAI say AI now writes 100% of their code.
If you're not using AI, you're spending 40 hours doing what they do in 4.
These 100+ Claude Code hacks fix that and help you ship 10x faster.
Sign up for The Code and get:
100+ Claude Code hacks used by top engineers — free
The Code newsletter — learn the latest AI tools, tips, and skills to code faster with AI in 5 minutes a day
Quote of the Week
Amateurs hack systems, professionals hack people. — Bruce Schneier
Reading List
How to transform your software development lifecycle with AI, stage by stage.
How we built certgrep.sh, a free regex search engine over certificate transparency, using an occurrence-only index, just-in-time hydration, and a Tantivy trigram index.
Music theory is derived from first principles using physics, arithmetic, and JavaScript, showing how frequencies lead to octaves, tuning systems, scales, chords, and chord progressions. Interactive Web Audio examples turn concepts like equal temperament, modes, harmony, tension, and notation into arrays, ratios, and code programmers can reason about.
The author prototypes SELF, an executable format that replaces ELF with SQLite so binaries, symbols, dependencies, and metadata become queryable and modifiable with SQL. The prototype can run SQLite database files directly, handle dynamic linking, bundle full dependency closures, and even pack an entire userland into one deduplicated database.
Modern systems make benchmarking surprisingly unreliable: the same code can produce different results because of JIT decisions, garbage collection, CPU cores, clock speeds, memory placement, and environmental conditions. The key lesson is to treat performance as a distribution and verify that benchmark runs are actually comparable before drawing conclusions.
After seven years with Rust, the author reimplements a JSONPath library in Zig and compares the languages across tooling, project structure, testing, programming style, and memory management. Zig feels fast, straightforward, and promising as a C successor, but manual allocation and ownership introduce friction and bugs that Rust prevents by design, while Zig’s ecosystem remains immature.
As AI agents automate more coding, Domain-Driven Design becomes more important because understanding the business domain, shaping the model, and agreeing on the design still require human judgment. The post shows how practices like design-before-code, ubiquitous language, bounded contexts, and deep domain knowledge help teams guide agents and review generated code effectively.
Turns a Casio VX-4 calculator with just 8 KB of RAM into a working Telnet BBS written in BASIC. The project uses RS232 and a Raspberry Pi Zero W for connectivity while keeping the calculator itself in control of sessions and messages.
The author built RollTab, an iOS app featuring a 125M-parameter transformer that autocompletes piano performances in real time directly on an iPhone. By representing each note with multiple embedded attributes instead of a multi-token sequence, the on-device model reaches ~108 notes/sec and produces better continuations after DPO fine-tuning.
A discussion on the various reasons why performance is slow, and which ones are unlikely to be affected by LLMs. Originally written as a counter-point to Dan Luu's 'There's no reason for software to be slow anymore'.
After Amazon software kept shutting down his Fire HD tablet, the author used four AI models over five months to find and exploit an unpatched kernel vulnerability and gain root access. The $266 experiment shows how AI models can collaborate on real security research, from discovering the vulnerability and debugging failed approaches to ultimately removing the protected Amazon software.
The author used more than 1,000 photos and Gaussian Splatting to create an interactive 3D reconstruction of New York City, with Claude orchestrating much of the workflow. Across 280 experiments and 74 AI sub-agents, the project highlights both the power and limitations of AI agents when tackling visual tasks that still require substantial human judgment.
Speculative programmatic tool calling (sPTC) speeds up AI agent harnesses by pre-launching tool calls from partially generated code instead of waiting for the model to finish generating. The approach overlaps tool execution with token generation and can also parallelize independent calls, acting like a lightweight JIT compiler for agent-generated programs.
This is a comprehensive guide to reinforcement learning for LLMs, starting from first principles and building through the policy-gradient algorithms used in modern LLM training. It covers RLHF, verifiable rewards, REINFORCE, PPO, and the newer RL methods being used to improve reasoning, agents, reliability, and efficiency.
Watch, Listen
Casey Muratori's talk at BSC 2026, delving into root of the root of all evil, premature optimization.
How do you raise awareness of performance concerns inside an organization? At what stages of a project is a performance expert most useful? When did software development lose its emphasis on performance in the first place? In this interview with Rico Mariani, we talk about precisely these questions. He spent much of his career at Microsoft trying to fix performance problems, both directly and indirectly. While working on the C# CLR, he coined the term “pit of success” to express the need for platform developers to design APIs such that non-experts will create software that performs well by default.
Learn how to build full-stack applications and autonomous AI workflows using Meta's Muse models and the Muse Code coding harness. Andrew Brown guides you through model architecture, structured outputs, search grounding, and programmatic API integrations with SDKs like LangChain. You will also master advanced developer workflows in the Muse Code CLI, including custom skills, MCP server setup, automated goal decomposition, and containerized deployments with Docker Compose.
Interesting Projects, Tools and Libraries
14MB foundation model for tiny devices; phones, wearables, smart home, and robots.
The open-source agent harness - the runtime layer that turns an LLM into a working agent.
match functions in binaries by what they do, not what their bytes look like. behavioral function fingerprinting via microexecution.
Hister is a private search engine for the pages you visit and the files you keep. It indexes their full contents so you can find information again from the web interface, terminal, or an AI assistant connected through MCP.
Clean up Claude 5's token vomit with a separate LLM. Save your tokens, Claude 5 is hopeless.
An experimental interface for building software by editing persistent pseudocode and synchronizing it with an AI-generated implementation.
Embedded single-file knowledge graph database with vector search and full-text search for AI/RAG apps.
PostgreSQL backup tool with Point-In-Time-Recovery and restore verification.
Apache Maka (Incubating) is a local-first AI agent workspace. Model messages, tool calls, tool results, permission decisions, and termination events are recorded as an append-only log.
The open-source AI IDE for Claude Code, Codex, OpenCode, and more. Run agents in parallel, locally or in the cloud, and build reusable workflows.
Open-source Stripe alternative for stablecoins. Checkout, subscriptions, and marketplace payouts. Stripe-compatible API. Your very own payments stack with no fees.
Python Weekly - A free weekly newsletter featuring the best hand curated news, articles, tools and libraries, new releases, jobs etc related to Python.
Founder Weekly - A free weekly newsletter for entrepreneurs featuring best curated content, must read articles, how to guides, tips and tricks, resources, events and more.