Research stack built for humans now looks wrong for AI scientists
AnneliesGamble · x · 2026-07-23
This post argues that the current research stack was built for humans, not AI scientists, and that a new infrastructure layer is needed.
Key points from the attached text:
- In the 1660s, scientists struggled to share discoveries through private letters, which were slow and often secretive.
- Henry Oldenburg helped turn that correspondence network into Philosophical Transactions of the Royal Society in 1665, one of the first scientific journals.
- That system made discoveries public, citable, and scrutinizable, and became the template for modern scientific publishing.
- Today, the paper remains the unit of knowledge, but the format is increasingly mismatched to AI-era workflows.
- Amber Liu argues the whole stack — PDF, arXiv, conferences, peer review, GitHub, Weights & Biases — is optimized for human speed, making it hard for AI scientists to build cumulative knowledge or collaborate.
The core claim is that AI research will need its own knowledge infrastructure, not just faster models.
Related event: Research Paradigm Shift: Academic Papers Inadequate for AI Scientists(3 posts)→
More from Infra
- Podcast: Micron's HBM roadmap, Meta's AI infra, and NVIDIA's supply chain lock-in — BenBajarin · 2026-10-27
- SGLang MegaKernel + PD Pushes the TCO-Latency Frontier on GB300, SemiAnalysis Benchmarks Show — AccBalanced · 2026-10-27
- Anthropic hits $16.3B H1 revenue as compute efficiency jumps 5.2x in 18 months — daniel_mac8 · 2026-10-27
- Tenstorrent runs Proteina protein model at high throughput, claims no HBM shortage — gajesh · 2026-10-27
- YC-backed Hopper launches voice-agent inference cloud: 80 ms TTFT at $0.30/1M tokens — ycombinator · 2026-10-27
- A100 rental prices still rising in 2026 per a16z data, Jevons paradox hits compute — VraserX · 2026-10-27