Exploring AI history by training models on pre-1931 text.

May 17, 2026
Weekly Newsletter

Weekly Newsletter ☕️

Welcome to the newsletter edition of May 17th, where you receive the best-voted websites of the past week.

📰 Top websites of the week

1. Talkie: Vintage 13B Language Model logo Talkie: Vintage 13B Language Model

Talkie trains vintage LMs on pre-1931 text to converse with the past. Researchers can compare vintage and modern models on historical tasks to study AI behavior and forecasting.

Visit website → Upvote & Comment →

2. DO_NOT_TRACK logo DO_NOT_TRACK

DO_NOT_TRACK proposes a single, standard environment variable to opt out of telemetry and tracking in CLI and TUI software. It aims to replace fragmented opt-out flags with a universal privacy switch for developers.

Visit website → Upvote & Comment →

3. The Year in Songs logo The Year in Songs

A year-round calendar of songs sorted by the dates they reference. Pick any day and hear every track that belongs to it.

Visit website → Upvote & Comment →

🎖 Sponsored websites

hiData.ai logo

hiData.ai

hiData: AI to clean, analyze & generate data reports via plain ta

Visit website →
AI Math Solver Powered by Math GPT Free logo

AI Math Solver Powered by Math GPT Free

Solve math problems with the GPT-4o—upload text or images for fre

Visit website →
ProjectManagementTools logo

ProjectManagementTools

Optimize your Projects with ProjectManagementTools

Visit website →

🗞 More awesome websites this week

4. TI-84 Evo Graphing Calculator logo TI-84 Evo Graphing Calculator

TI-84 Evo is the most advanced graphing calculator in the TI-84 series, ideal for algebra, geometry, calculus, and other secondary-school math. It features a faster processor, larger graphing area, an icon-based home screen, and a redesigned keypad for quicker use, built for classrooms and exams with durable hardware.

Visit website → Upvote & Comment →

5. NetHack 5.0.0 Release Notes logo NetHack 5.0.0 Release Notes

NetHack 5.0.0 release notes document game improvements, architectural changes, and cross-compiling enhancements. The notes explain the switch from legacy tools to Lua-based processing, list over 3100 fixes, and provide verification commands, compatibility caveats, and bug reporting guidance.

Visit website → Upvote & Comment →

6. You Can Beat the Binary Search logo You Can Beat the Binary Search

Daniel Lemire's performance analysis demonstrates how to beat the classic binary search algorithm. The post explores how modern CPU architecture and memory access patterns make quaternary search faster and more efficient in practice.

Visit website → Upvote & Comment →

by @Micadep