DSpark: Speculative decoding accelerates LLM inference [pdf]
783 pts, 361 comments
783 points · 361 comments · Add your take →
Top Hacker News Stories – June 27, 2026 (Summarized)
Recent discussions on Hacker News highlight significant advancements in large language model (LLM) inference, with DSpark's speculative decoding technique gaining notable attention, earning 783 points and over 360 comments. Meanwhile, Apple’s Neural Engine architecture and performance analysis attracted 232 points, reflecting ongoing interest in hardware optimization. In addition, Reddit’s anti-spam system, which employs shadowbans and a second-chance pool, was examined for its internal mechanisms, sparking further debate. Other stories include the release of OpenTTD 16.0-Beta1 and a nostalgic look at the classic WordStar word processor from 1996.
Free the Icons
680 pts, 257 comments
680 points · 257 comments · Add your take →
Apple Neural Engine: Architecture, Programming, and Performance
232 pts, 29 comments
232 points · 29 comments · Add your take →
OpenTTD 16.0-Beta1
232 pts, 46 comments
232 points · 46 comments · Add your take →
A peek into Reddit's anti-spam internals
Reddit's anti-spam system involves shadowbans and a second-chance pool, causing some posts to reappear.
185 points · 77 comments · Add your take →
WordStar: A Writer's Word Processor (1996)
173 pts, 92 comments
173 points · 92 comments · Add your take →
Anatomy of a Failed (Nation-State?) Attack
155 pts, 33 comments
155 points · 33 comments · Add your take →
Adrafinil – keep a lid-closed Mac awake only while agents work
124 pts, 78 comments
124 points · 78 comments · Add your take →
Old Computer Challenge
116 pts, 79 comments
116 points · 79 comments · Add your take →
Hacker News on a train station-style flip board
114 pts, 30 comments
114 points · 30 comments · Add your take →
Alan Kay on the meaning of "object-oriented programming" (2003)
91 pts, 91 comments
91 points · 91 comments · Add your take →
Kb – Prolog Knowledge Base
84 pts, 18 comments
84 points · 18 comments · Add your take →
Software Is Becoming Marketing
67 pts, 32 comments
67 points · 32 comments · Add your take →
Is America becoming a gerontocracy?
55 pts, 111 comments
55 points · 111 comments · Add your take →
IBM MCGA Gate Array Reverse Engineering
53 pts, 8 comments
53 points · 8 comments · Add your take →
Netflix Simplified Batch Compute with Kueue
48 points · 7 comments · Add your take →
Starglyphs - A constellation puzzle game based on Euler paths
26 points · 7 comments · Add your take →
MacBook vs. Dedicated GPU for LLM
22 points · 71 comments · Add your take →
Kiso, an open-source publishing engine for Open Knowledge Format
19 points · 0 comments · Add your take →
AI Erodes a Legacy of Reading
10 points · 4 comments · Add your take →
They quit the West for Russia's traditional values, but it wasn't as expected
9 points · 2 comments · Add your take →
Ukrainian Attacks Spur State of Emergency Declaration in Crimea
9 points · 0 comments · Add your take →
NLNet Labs LLM Policy
9 points · 2 comments · Add your take →
Decline in IQ linked to frequent or dependent use of cannabis use in youth
8 points · 4 comments · Add your take →
Walter S. Arnold–Sculptor/Stone Carver
8 points · 2 comments · Add your take →
KV-psi, using Linux PSI to to trim an LLM KV cache
8 points · 0 comments · Add your take →
A Living Neural Web in HTML5 Canvas
6 points · 6 comments · Add your take →
The TypeScript Semantic Layer for ClickHouse
6 points · 6 comments · Add your take →
I got tired of spending 3hrs daily on job applications,so I automated it
5 points · 3 comments · Add your take →
Iterating Faster with TypeScript 7
5 points · 3 comments · Add your take →
Russian hackers were behind $2.5B hack of Jaguar Land Rover
5 points · 0 comments · Add your take →
Get weekly best stories
One email with the best technology stories and topic pages. No spam.