Ask HN: Why is LLM training still GPU-hungry despite DeepSeek?

1 takinola 0 7/15/2025, 2:25:19 PM

When DeepSeek released R-1 everyone thought that signaled the end of the GPU-intensive LLM training approach. It does not appear to have worked out that way as GPU demand continues to grow unabated. What happened? Is the DeepSeek training method unreproducible or impractical in some way?

Syndrome of populism and the evolution of post-Soviet regimes (frontiersin.org)

When you look at the top source of attacks (ASN), Hetzner is at a 3rd place (blog.cloudflare.com)

Ask HN: Why is LLM training still GPU-hungry despite DeepSeek?

Data breach that put up to 100k lives at risk revealed (independent.co.uk)

Two men behind 'senseless' felling of Sycamore Gap tree jailed for four years (theguardian.com)

Is your company invisible to ChatGPT, Gemini, and Claude? (ismycompanyonai.com)

Quick notes from the GStreamer Spring Hackfest 2025 (collabora.com)

Practical Engineering: How Floating Bridges Work [video] (youtube.com)

Chrome OS is "combining" with Android, but what does that mean? (arstechnica.com)

First Feedback Filter–reducing regret in responding to negative feedback (tidyfirst.substack.com)

Postalon: The AI agent for shitposting on X.com (postalon.vercel.app)

A foundation model to predict and capture human cognition (nature.com)

Setup Postiz on Coolify (softuts.com)

Dumbest Mistakes We Made Migrating from Google Analytics to Snowplow (edistel.substack.com)

John Jumper: AlphaFold and the Future of Science [video] (youtube.com)

China's Baidu to bring its driverless cars to Uber globally (cnbc.com)

Show HN: PosFast – Not Vibe Coded Social Media Scheduler (postfa.st)

Why Did Y Combinator Delete "DocuSign 2.0" from Its RFS? (unicornforms.com)

Show HN: RAGsplain – What does your RAG model see before it answers? (ragsplain.com)

The next generation of managers must be fit for a net-zero nature driven economy (nature.com)

Half of homeless people have experienced traumatic brain injury: study (globalnews.ca)

Power prices are expected to soar under new tax cut and spending law (npr.org)

The Plato Plateau – anxiety matched philosophy (aneeshsathe.com)

Headscale: Self-hosted implementation of the Tailscale control server (headscale.net)

What birdsong and back ends can teach us about magic (digitalseams.com)

Layout Maestro CSS Course (layoutmaestro.ishadeed.com)

Show HN: I made Instant Resume to make your resume in minutes (instantresume.online)

Tiptap 3.0 is stable: open-source text editor built on ProseMirror (tiptap.dev)

The Rushed Development of MiniDisc (obsoletesony.substack.com)

ABCD Study omits gender-identity data from latest release (thetransmitter.org)

Thousands offered UK asylum in secret scheme after personal data of Afghans (theguardian.com)

Dark Matter-Powered Objects Awaiting Discovery at the Galactic Center (arxiv.org)

Elon Musk's AI predictions should terrify us (damianreilly.co.uk)

A New Record in Self-Cleaning Turing Machines (nickdrozd.github.io)

Show HN: Fertit – Open-Source Newsletter Manager with Optional Hosted Service (fertit.com)

Harnessing Frustration: Using LLMs to Overcome Activation Energy (lalitm.com)

Milk-V Titan Brings RISC-V Performance in Mini-ITX Form with UltraRISC Ur-DP1000 (linuxgizmos.com)

Record Labels: A "Safe Haven for Pirates" Disqualifies ISP from DMCA Protection (torrentfreak.com)

OpenZFS Bug Ported to C (flak.tedunangst.com)

Commodore 64 Is Back–and More Gamer-Fueled Than Ever with a Transparent RGB Case (gizmodo.com)

Incel language infected the mainstream internet (theverge.com)

Why Women Face Higher Alzheimer's Risk (neurosciencenews.com)

Fixing the R&D Tax Code Blunder Isn't a Victory, It's a Reset (news.bloombergtax.com)

Animal revenge: Can tigers, orcas, and other animals get back at their enemies? (slate.com)

More Artificial than Intelligent, and it is only getting worse (mlagerberg.com)

Misleading macOS “App is Damaged” Warning (github.com)

Chip4Mac68000 – A Chip8 emulator for the original Macintosh (github.com)

Show HN: BYOK Node Based Image Gen Playground with OAI, Flux, Geminin (hitslop.com)

Parsing 1B Rows in Bun/TypeScript Under 10 Seconds (taekim.dev)

Show HN: Zzz – IDE for Kotlin mobile development on Windows (s0mbra.com)

Ask HN: Why is LLM training still GPU-hungry despite DeepSeek?

Comments (0)