Hacker News Digest

Wednesday, May 20, 2026

In This Issue

  • Hacker News
  • Andrej Karpathy joins Anthropic
  • Gemini 3.5 Flash
  • I've built a virtual museum with nearly every operating system you can think of
  • Minnesota becomes first state to ban prediction markets
  • New accessibility features powered by Apple Intelligence
  • Google changes its search box
  • Graduates are booing pep talks on AI at college commencements
  • Google Search as you know it is over
  • AI is too expensive

Zipper Data Brief

May 20, 2026
Your daily digest of the best from Hacker News

Top 6 Trending

#1
1353 points · dmarcos · comments
Discussion Summary
Andrej Karpathy joining Anthropic's pre-training team is seen as a major hire that signals both Anthropic's strength and concerns about AI consolidation, though many worry it may reduce his valuable educational output and represents a shift from independent innovation to working within frontier labs.
#2
835 points · spectraldrift · comments
Discussion Summary
Gemini 3.5 Flash is fast and capable for a flash model, but Google has tripled pricing compared to previous versions, raising concerns about whether the industry is quietly resetting inference costs upward rather than making AI cheaper as promised. Performance improvements don't clearly justify the price hike, and some benchmarks show it underperforming cheaper alternatives like open-source models and previous Google releases.
#3
824 points · andreww591 · comments
Discussion Summary
Users praise this ambitious OS museum project for its preservation work, but note missing systems (Pick, BeOS, TempleOS, Novell Netware), suggest improvements like searchability and downloadable lists, and point out that emulation captures visual appearance rather than the tactile experience that defined these systems.
#4
688 points · ortusdux · comments
Discussion Summary
The discussion centers on whether Minnesota's ban on prediction markets is legally enforceable given federal CFTC jurisdiction over futures markets, with commenters debating the merits of the ban itself—some viewing prediction markets as harmful gambling that should be restricted, while others question the state's authority and worry about overreach like banning VPNs.
#5
685 points · interpol_p · comments
Discussion Summary
The discussion praises Apple's genuine accessibility investments as genuinely useful, particularly on-device AI features, while blind developers and users temper enthusiasm by noting existing third-party alternatives, maintenance gaps in VoiceOver, and emphasizing the critical need to involve disabled people in product development rather than building solutions in isolation.
#6
583 points · berkeleyjunk · comments
Discussion Summary
Users express deep concern that Google's shift toward AI-generated summaries will destroy web traffic for small sites, reduce access to primary sources, enable undisclosed advertising, and ultimately harm the open internet—while ironically creating opportunities for competitors like Kagi and DuckDuckGo.

AI / Machine Learning

116 points · 1vuio0pswjnm7 · comments
Discussion Summary
Graduates booed AI-focused commencement speakers because tech executives promoting AI automation ignored the real economic anxieties of young people facing job displacement, student debt, and an uncertain future—while displaying detachment from how their optimistic rhetoric sounded tone-deaf to those bearing the actual costs.
111 points · evo_9 · comments
Discussion Summary
Users are frustrated that Google Search's new AI-generated answer feature is inferior to traditional search results, producing hallucinations and killing traffic to content sites, while raising questions about Google's unsustainable cost model and the broader collapse of ad-supported media that feeds the AI training data.
139 points · crescit_eundo · comments
Discussion Summary
The discussion centers on whether massive AI investments will eventually become profitable or represent a bubble, with commenters divided between those believing breakthrough efficiency gains and killer use cases justify current spending, and skeptics arguing costs are unsustainable and pricing will eventually spike once users become dependent.
55 points · maxheyer · comments
Discussion Summary
The discussion centers on a fundamental skepticism that abandoning hands-on coding for AI code review risks atrophying programming skills and narrowing problem-solving ability, with commenters emphasizing that understanding code through doing—not just reviewing—is essential to effective engineering.
181 points · thanhhaimai · comments
Discussion Summary
Commenters criticized Google I/O for being AI-focused and underwhelming, with presentations lacking the innovation and polish of past tech events, while expressing concerns about pricing, feature availability locked behind "trusted testers," and Google re-releasing existing products rather than pioneering new ones.

Startups / Business

32 points · walkervin · comments
Discussion Summary
A solo developer apparently disrupted YouTube's "Ask" feature, with commenters expressing skepticism about the claim and questioning both its feasibility and the suspicious presence of new accounts promoting it.
274 points · doener · comments
Discussion Summary
Mistral AI's acquisition of Emmi AI to build an industrial AI stack is seen as a smart vertical differentiation strategy, though some remain skeptical about the company's competitiveness against the Big 3 and question whether Europe can sustain AI leadership given capital, regulatory, and brain drain challenges.
85 points · vincent_s · comments
Discussion Summary
The discussion reflects skepticism about Anthropic's IPO plans, with commenters concerned about unsustainable valuations, the lack of clear profitability paths, and the likelihood that consumers will face significant price increases once the company goes public and needs to recoup investments.
17 points · kpw94 · comments
Discussion Summary
A commenter makes a quip suggesting Anthropic has massive funding to spend, likely referencing recent capital raises in the AI industry.
8 points · swolpers · comments
Discussion Summary
A commenter skeptically suggests Anthropic will decrease safety oversight while claiming the reduction represents progress.

More Stories (34)

309 points · pabs3 · comments
416 points · 7777777phil · comments
399 points · bradley_taunt · comments
57 points · Larrikin · comments
314 points · janalsncm · comments
Alternatives to HN for "tech outside of AI" discussion?
28 points · summonerOS · comments
26 points · Dinux · comments
92 points · kgraves · comments
255 points · claaams · comments
Created by Zipper Data Co.  · 2026-05-20 12:01 UTC  · Unsubscribe

Get digests like this delivered to your inbox every morning.

Subscribe Free