Click any tag below to further narrow down your results
Links
A developer walks through how their daily coding work changed across ChatGPT, Cursor tab completion, Claude agents, and Opus 4.5 through 5.5, ending with a model building a working notes app for iPhone from a single prompt. It reads as a personal timeline of AI tools moving from a better search box to something that handles most of the implementation.
- Productivity rose from a modest gain to roughly 3x (measured over a few months of merged PRs) once the Opus 4.5/6 models needed far less step-by-step prompting.
- The developer's role shifted from writing and reviewing every line to focusing on architecture, business logic, customer data, and irreversible changes, scanning the rest.
- A single prompt produced a basic working iPhone app that imported all their Electron notes, using about 40% of a $20 subscription's 5-hour limit.
AI coding tools aren't producing blockbuster apps, but they're driving a 2x boom in modestly successful niche software—game mods, browser extensions, plugins—that serve hundreds or thousands of users. The author analyzed installation counts across 20 software marketplaces and found clear growth in small projects with no corresponding explosion in major software.
- Between 2022-24 and 2025-26, successful small software projects roughly doubled (plugins, mods, extensions, game add-ons), while massively successful software showed no growth.
- Big software companies aren't shipping more features because they already have what users want and are technically constrained by existing complexity; new wildly successful projects take years to develop.
- Most hobbyist developers won't publicly admit using AI due to community backlash, making voluntary disclosure data unreliable—the 2x growth in projects against 10x growth in total projects is the real signal.
Anthropic built Scion, a custom version control system, to replace Git because Claude's code generation capability pushed their merge rate from 2,500 to 30,000+ code changes per day—a scale Git's distributed design couldn't handle. The system treats clients as always-online agents and moves computation to the server, ditching Git's model of complete local repositories.
- Anthropic's monorepo became unmanageable under Git: the history alone is 100 GB, and they were doing 25 million checkouts daily for CI, which would generate 600 PB of traffic if done naively.
- Scion's core design flips Git's assumptions: clients stay thin and online, the server handles all history queries and merges, and there's no single fully-materialized copy of the repo anywhere—everything is distributed and horizontally scalable.
- The system uses "changesets" (write-ahead log entries) instead of Git commits, storing only what changed relative to the parent plus content-addressed pointers to blob storage, allowing high-throughput acceptance with no packfiles or cross-server locking needed.
This article traces the evolution of AI loops—small programs that run, check, and re-prompt coding agents—from early ReAct and AutoGPT examples to today’s durable, multi-agent orchestration with scheduling and self-verification. It shows why loop management, not model calls, is now the biggest cost in AI coding and outlines best practices: cap iterations, build reusable skills, and include feedback checkpoints.
- Boris Cherny landed 259 PRs in 30 days by having loops handle all the prompting instead of him
- Loops have evolved through five stages, from ReAct/AutoGPT to today's self-scheduling, crash-recoverable multi-agent systems
- The real difference from plain cronjobs is a built-in decision engine: loops read state, decide next steps, and validate results rather than just firing a fixed script
- Cherny's practical advice centers on self-verification, auto-permissions, and using /loop or /goal commands so agents catch their own mistakes
Anthropic released a two-hour course led by the engineer behind Claude Code that walks you through building self-managing Claude agents. It covers terminal integration, file-system memory, hallucination-blocking hooks, and scaling to large codebases. Whether you’re a beginner or advanced user, you’ll finish ready to use Claude professionally.
- Anthropic released a free two-hour course on building Claude agents, led by the engineer who writes Claude Code
- Covers terminal integration, file-system-based memory, and hooks designed to catch and block hallucinations
- Includes guidance on scaling agents to large codebases via isolated environments and multi-file task orchestration
- Shows real failure examples (missing dependencies, misread logs) and how to harden pipelines against them
OpenAI has launched GPT-5.1-Codex-Max, a new coding model designed to enhance agentic tasks in software engineering. This model features improved speed, token efficiency, and the ability to manage long-running tasks by compacting context windows, positioning it as a more reliable coding partner for developers.
- GPT-5.1-Codex-Max is trained natively for agentic coding across multi-hour, multi-step tasks via "compaction," letting it clear context and keep working across millions of tokens rather than hitting a fixed window limit
- Positioned as faster and more token-efficient than prior Codex models while maintaining coherence over long-running sessions
- Marketed as a more reliable long-horizon coding partner for real-world software engineering workflows, not just short code completions
Boris and Cat, the team behind Claude Code, explain how a terminal tool grew out of an internal prototype after an engineer was told to use it instead of hand-writing code. They cover why giving the model bash and a few tools worked better than building a custom toolkit.
- Claude Code began as an internal research tool called Clyde. A new engineer's hand-written PR got rejected with "use Clyde," and the one-shot result convinced the team that an IDE might not be necessary.
- The turning point was giving the model bash. It started writing AppleScript and automating things on its own, which the team had not expected.
- Anthropic "dogfoods" the product: roughly 70 to 80 percent of technical staff use it daily, and internal feedback arrives constantly, so features get shipped or pulled within days.
- The team has been removing tools over time. The LS tool was dropped once bash permissions could enforce directory restrictions, which leaves less clutter in the model's context.