Article URL: https://rorz.io/writing/my-friend-aaron Comments URL: https://news.ycombinator.com/item?id=49437069 Points: 7 # Comments: 1
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
A solo developer building 'Balance', a portfolio rebalancing app, engineered a single core engine to support three distinct markets — Brazil's B3, US stocks, and cryptocurrency — rather than building separate systems for each. The app uses a single 'market' field on the Portfolio model to branch only where necessary, such as price sources (BRAPI, Yahoo Finance, Binance), currency symbols, tax rules, and quantity precision. Approximately 90% of the domain logic, including rebalancing math, deposit flows, and snapshots, remains fully shared across all three markets. A key technical challenge was decimal precision: crypto quantities require up to 8 decimal places, while stocks use whole numbers, solved by storing all values at maximum precision and rendering them differently in the UI. The design means adding a new market requires changes only at specific, isolated branch points, leaving the core rebalancing pipeline untouched.
A developer named Drew has built an open-source tool called 'Simplify Your Writing' that helps users replace wordy phrases with concise alternatives. The tool runs entirely in the browser using a predefined dictionary of phrase-to-replacement rules, with no AI involved, making its suggestions transparent and predictable. Unlike AI writing assistants, it does not rewrite text or alter the author's voice — it only flags specific phrases and lets the user decide whether to change them. The project is currently seeking contributors to expand its dictionary across domains such as technical, academic, business, and everyday writing. Developers can also help improve the matching engine, UI, and accessibility via the project's GitHub repository.
Retrieval-Augmented Generation (RAG) prototypes often perform well in demos because developers hand-pick clean, consistent documents for testing. When the same systems are deployed on real organizational data, they encounter contradictory policies, outdated files, and poorly formatted content, causing AI responses to be confidently wrong. Unlike a search engine that surfaces bad documents for human review, an AI assistant buries flawed source material inside fluent, authoritative-sounding prose. Practitioners argue the most critical early step is not writing code but mapping which documents actually govern each topic, who owns them, and what should be excluded from the index entirely. Many RAG projects also fail because the answers users need exist in Slack threads or institutional memory rather than any indexed document, making pre-build content audits essential.
GitHub overhauled its Copilot billing system on June 1, 2026, replacing premium request-based metering with a new GitHub AI Credits model priced at one cent per credit. Under the old system, each paid plan included a fixed monthly allowance of premium requests per user, ranging from 50 for free-tier users to 1,500 for Pro+ subscribers. Advanced features like code review and agent mode consumed requests at varying multipliers depending on the AI model used, meaning two users with similar activity could deplete their allowances at very different rates. Annual Copilot Pro and Pro+ subscribers who have not switched plans continue to be billed under the legacy premium request framework. GitHub has officially labelled premium requests as 'legacy' in its documentation, though understanding the old system remains relevant for users transitioning to the credits model.
Discussion (0)
Log in to join the discussion and vote.
Log in