Firefox is the only major browser still supporting uBlock Origin as others switch to Manifest V3.
Two camps: those cheering Firefox's newfound ad-blocking edge, and those doubting it matters enough to shift users.
1328 pts · 507 comments
Article body wasn't reachable. HN discussion still summarized.
Debate over whether the 27B model truly beats Opus on real-world usage versus benchmarks.
1234 pts · 726 comments
Scaling post-training on GLM-5.3 produces gains in coding and emergent cyber exploitation capabilities.
A focus on open-weight unrestricted models versus locked-down US frontier models for cybersecurity work.
1111 pts · 545 comments
Opus 5 is more capable but feels worse to use due to unwanted autonomy.
Many users report Opus 5 cheats, writes cryptic comments, and is slower than earlier versions.
901 pts · 808 comments
A satirical page recreates every annoying website UX pattern from 2020.
Commenters agree the satire is accurate, sharing UX pet peeves and debating EU cookie law blame.
810 pts · 468 comments
Count Binface received 26.9% of votes in Clacton by-election, his best ever result.
Binface got more votes than needed to trigger a recall petition against Farage.
446 pts · 367 comments
Google's open-source HEIR compiler enables practical private AI inference via homomorphic encryption.
Two camps: cryptography believers versus skeptics of Google's trust and motives.
441 pts · 262 comments
AI-powered bug hunting could eliminate remotely-exploitable vulnerabilities, driving law enforcement to demand backdoors.
Three camps: those who welcome reduced state hacking, those who fear backdoor mandates, and skeptics of the premise.
388 pts · 186 comments
A writer keeps seven beloved books on an elbow-high shelf for inspiration while writing.
Roget's original classification structure is praised and contrasted with modern synonym-list editions.
355 pts · 163 comments
Australia's home battery subsidy program cut wholesale power prices by 47%.
Two camps: use EVs as storage vs. home batteries; cycle counts and V2G limits debated.
342 pts · 299 comments
Prof. Tom Yeh teaches AI math, algorithms, and architectures through hand-drawn diagrams.
Frustration with the paywall overshadowed enthusiasm for the educational content.
336 pts · 25 comments
RustDesk offers true unattended remote access on Wayland with multi-monitor support.
Debate over RustDesk's password security and license status split the thread.
322 pts · 133 comments
Qwen3.8-27B
(X (formerly Twitter))
Alibaba releases open-weight Qwen3.8-27B model outperforming Qwen3.7-Plus.
User compares Qwen3.8-27B to Qwen3.6-27B for agentic workflows.
297 pts · 3 comments
Claude Code's efficiency depends on managing prompt caching, context size, and session structure.
Unexpected cache rewrites inflate costs to over $100 per session for some users.
271 pts · 153 comments
Newly disclosed records show DHS spied on leftwing groups and anti-ICE protesters via undercover agents.
Historical parallels to COINTELPRO dominate, with debate shifting to practical countermeasures like compartmentalization and encryption.
263 pts · 93 comments
DeepSeek launched V4-Pro with agent upgrades and peak/off-peak API pricing.
Duplicate submission flagged; comments redirected to the main thread.
237 pts · 3 comments
Article body wasn't reachable. HN discussion still summarized.
Split between device-level enforcement advocates and those citing privacy risks.
223 pts · 164 comments
Toast 1 is a specialized search agent matching frontier model quality at lower cost and latency.
Skepticism over the joke-sounding brand name, mixed with questions about how Toast 1 differs from standard RAG.
211 pts · 64 comments
A developer built a tool to turn RSS feeds into e-pubs for a pocket e-reader.
A range of e-reader RSS solutions exist, but phone dependency for banking remains a barrier for some.
209 pts · 90 comments