SShortSingh.
Back to feed

Non-coder builds AI companion app leaderboard using Claude Code and vibe coding

0
·1 views

A self-described non-coder built a community-voted leaderboard for AI companion apps using Claude Code, without writing a single line of code manually. The platform hosts seven leaderboards with monthly snapshots, built on Astro, Cloudflare Pages, and D1, addressing the lack of transparent, unbiased rankings in the AI companion app space. To counter potential fake votes from affiliate-incentivized apps, the system silently accepts over-limit ballots but routes them to a shadow table excluded from scoring, preserving a full audit trail. A key lesson shared was the risk of renaming UI labels that are silently tied to API keys, which can break functionality without any visible error. The leaderboards are live at nofilterreview.com/rankings, with plans to add an automated agent that scores apps on memory and character consistency.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer builds free structured learning platform, seeks collaborators to scale it

A developer built a free learning platform called Learn by Play after growing frustrated with fragmented, paywalled educational resources online. The platform currently offers structured content for both spoken language learning and programming concepts, with voice audio support being actively added. Built entirely in the developer's spare time, the site is already live and accessible to the public. The creator is now seeking collaborators with frontend, backend, or content design skills to help grow the project beyond a solo effort.

0
ProgrammingDEV Community ·

Developer Launches NAEOS to Bridge the Gap Between AI Tools and Software Engineering

A developer has introduced NAEOS, short for Nusantara AI Engineering Operating System, as a foundational layer designed to sit between developers and AI coding agents. The project stems from the founder's repeated experience of rebuilding the same engineering groundwork — architecture, coding standards, documentation, and workflows — at the start of every new software project. While tools like GitHub Copilot, Cursor, and Claude Code excel at generating code, the founder argues they lack persistent engineering context, governance, and architectural understanding. NAEOS aims to fill that gap by providing reusable engineering systems rather than relying solely on prompt engineering. The goal is to enable AI agents to build software the way experienced engineering teams would, with consistent standards and long-term maintainability.

0
ProgrammingDEV Community ·

Developer Builds 6.4M-Parameter Transformer From Scratch to Power Recipe App

A developer built a custom decoder-only transformer called RasavedaGPT entirely from scratch using PyTorch, without relying on any external APIs or pretrained weights. The 6.4-million-parameter model powers Rasaveda, a recipe intelligence app that matches ingredients to recipes and explains cooking steps conversationally. Trained in two stages on a single Google Colab T4 GPU, the model first learned general English from WikiText-2 before being fine-tuned on recipe-specific data. At its small size, the model runs inference on CPU within a FastAPI backend, eliminating API costs, rate limits, and third-party dependencies. The project was driven by both practical self-sufficiency goals and a personal desire to deeply understand transformer architecture from the ground up.

0
ProgrammingDEV Community ·

Developer builds Astra, a modular Unix shell written in Rust

A developer has released an early-stage personal project called Astra, an interactive command-line shell built using the Rust programming language for Unix-like systems, with macOS as the primary test platform. The shell is designed around a clean, modular architecture that aims to be easy to understand, extend, and contribute to. Current features include a customizable prompt, TOML-based configuration, built-in themes, Git-aware prompts, command history, tab completion, and alias support. A plugin framework is also in early development alongside documentation, packaging, and GitHub automation to help onboard contributors. The project is not yet stable, and the developer advises against using it as a primary shell until a stable release is issued.