Releases·Tool Release
Developer ships aimake 2.0, an open-source incremental build tool that caches and reruns only changed AI pipeline steps
A developer released aimake 2.0, an open-source incremental build system for AI/ML pipelines, on GitHub. It builds a dependency graph over pipeline stages (data, preprocessing, model, embeddings, prompts, evals, reports) and uses SHA-256 content fingerprints to rerun only what changed. The tool targets the common pain of rerunning entire pipelines after a single-stage change.
1 engagement1 sourceThu, Sep 3, 2026, 04:40 PM
AI Overview
- aimake 2.0 builds dependency graphs with SHA-256 content fingerprints to reuse cached steps and offers plan/build/explain CLIs, cost estimation, and S3 remote caching
- aimake ships Python and TypeScript SDKs, a TUI and web dashboard, and plugins for Hugging Face, W&B, DVC, Docker, and Ollama
Related
Products·launched
aimake — incremental build system for AI/ML and agent pipelines
aimake is an open-source incremental build system, like make for AI apps. It models pipelines (dataset, preprocessing, embeddings, prompt, eval) as dependency graphs, fingerprints inputs by content, and rebuilds only affected steps. Built with Claude Code, it targets developers of AI/RAG pipelines who want to avoid unnecessary reruns.
aimake is an open-source incremental build system, like make for AI apps. It models pipelines (dataset, preprocessing, embeddings, prompt, eval) as dependency graphs, fingerprints inputs by content, and rebuilds only affected steps. Built with Claude Code, it targets developers of AI/RAG pipelines who want to avoid unnecessary reruns.
11012 engagement3 sourcesredditTue, Sep 1, 2026, 06:16 PM
Releases·Tool Release
Revision Prompting: A trick to avoid regenerating the whole output when only 10% of the input changed.
A developer shares a technique called revision prompting for LLM pipelines: when input changes slightly, instead of re-running the full prompt, send the old input/output plus a diff and ask for a patch to the output. This reduces cost and keeps unchanged parts identical. The write-up is at revisionprompting.info.
A developer shares a technique called revision prompting for LLM pipelines: when input changes slightly, instead of re-running the full prompt, send the old input/output plus a diff and ask for a patch to the output. This reduces cost and keeps unchanged parts identical. The write-up is at revisionprompting.info.
31 engagement1 sourceredditSun, Aug 9, 2026, 02:02 PM
Releases·Tool Release
Revision Prompting: Trades slow (decoded) output tokens for cheap (prefilled) input tokens.
A technique called revision prompting reduces output token generation by 2-10x in repeated-prompt pipelines by sending the old input/output plus a diff and asking for a patch. It exploits the fact that prefill is cheaper than decode, keeping untouched output byte-identical.
A technique called revision prompting reduces output token generation by 2-10x in repeated-prompt pipelines by sending the old input/output plus a diff and asking for a patch. It exploits the fact that prefill is cheaper than decode, keeping untouched output byte-identical.
24 engagement1 sourceredditTue, Aug 11, 2026, 12:22 PM
Products·launched
npm package to compress project context for switching AI tools
A developer tool that gathers a project's codebase into a compact context and compresses it, allowing users to switch between AI coding tools without losing project context. It reduces repeated setup and costs by avoiding overlapping tool subscriptions.
A developer tool that gathers a project's codebase into a compact context and compresses it, allowing users to switch between AI coding tools without losing project context. It reduces repeated setup and costs by avoiding overlapping tool subscriptions.
5 engagement1 sourceredditFri, Aug 21, 2026, 09:31 AM
Products·launched
AICurt: Lightweight AI text pre-processing toolkit for LLM workflows
AICurt is a lightweight Python toolkit that simplifies text pre-processing for AI workflows, including steps needed before sending data to LLMs, embedding systems, or retrieval pipelines. It solves the problem of repetitive, complex data preparation for developers building AI applications.
AICurt is a lightweight Python toolkit that simplifies text pre-processing for AI workflows, including steps needed before sending data to LLMs, embedding systems, or retrieval pipelines. It solves the problem of repetitive, complex data preparation for developers building AI applications.
0 engagement1 sourcerssSat, Jul 25, 2026, 07:36 AM