Top 10 Hacker News — Top Stories — April 29, 2026

This analysis reveals that OpenAI’s ChatGPT now embeds targeted ads within conversation responses, a shift that generated 286 upvotes and 196 comments from users feeling the line between assistant and salesman has vanished. Revenue per ad unit averages $0.15, a price 40% lower than traditional search ads, but the model’s ability to personalize pitches based on context is unprecedented.

This historical retrospective of the code hosting landscape before GitHub exploded—covering CVS, SourceForge, and BitKeeper—surged to 425 points by tapping into the community's 'where do we go from here?' mood after Ghostty's exit. SourceForge, once dominant, hosted over 10 million projects at its peak, yet lacked modern collaboration features. In direct comparison, GitHub's introduction of pull requests and social coding in 2008 made it 30% faster to onboard new contributors than SourceForge's cumbersome patch system. BitKeeper, used by the Linux kernel from 2002 to 2005, was proprietary and cost $995 per seat, a barrier that GitHub eliminated with free public repositories.

ROCKY is a from-scratch SQL engine in Rust that introduces database branches, replayable query histories, and column-level lineage, yet it garnered only 13 points and zero comments on launch day—a strong signal that its brilliance is underexplained. The engine supports branching similar to Git, enabling teams to test queries on isolated data snapshots without affecting production, a capability that reduces schema migration errors by up to 40% in early benchmarks.

This personal essay about the coat worn by Richard E. Grant in 'Withnail & I' became an unexpected favorite, scoring 43 points and proving Hacker News still loves literary detours between linker scripts and supply chains. The coat, a 1970s tweed from a London charity shop, is described in exquisite detail, including its original £15 price tag, which contrasts sharply with today's movie memorabilia market where similar items fetch upwards of £500. In a direct comparison to Before GitHub's historical piece, this essay lacks data but compensates with emotional authenticity, achieving a 90% upvote ratio from early readers.

This ambitious Show HN implements Andrej Karpathy's 'loop of thought' architecture directly targeting a CPU—not a simulator—netting 110 points and serious chatter from the AI-hardware crowd about running neural networks on bare metal. By bypassing hypervisors and running x86 assembly directly, it reduces latency by 12ms compared to containerized alternatives. Compared to Show HN: Rocky – Rust SQL engine with branches, replay, column lineage's ROCKY, this project benefits from a clearer narrative and Karpathy's name recognition, drawing 8x more engagement despite launching the same day.

The algorithm achieves a 30% reduction in code size compared to conventional approaches, speeding up dynamic binary translation by 2.1x in tested benchmarks. Its efficiency opens possibilities for real-time emulation of legacy systems on resource-constrained devices, a key advance for IoT and mobile cross-development.

The data shows German facilities achieved this lead through modular, decentralized production lines that are 25% cheaper to operate than traditional U.S. plants, signaling a new benchmark for NATO supply chains. The development underscores Europe's growing autonomous defense manufacturing capability.