digestdaily-digestai-newsftl-lab
🌅 FTL Lab — AI Daily Digest · October 01, 2026
Today: 10 new articles, 5 trending models, 5 research papers
Data Pulse
- 10 news articles
- 5 trending models
- 5 research papers
- Cheapest GPU: Tesla V100 at $0.03/hr
Top AI Articles Today
Google made its boldest bid yet for AI dominance with the release of Gemini 4 Argon, while OpenAI escalated its fight against model theft and expanded its push into small-business adoption. Meanwhile, the money kept flowing — ElevenLabs doubled its valuation to $22 billion, and a hardware-design startup hit $750 million — even as a bitter spat between Factory and Cognition laid bare just how cutthroat the talent and IP wars have become.
- NVIDIA Opens Applications for 2027–2028 Graduate Fellowships With Awards Up to $60,000 — NVIDIA has opened applications for its 2027–2028 graduate fellowship program, offering awards of up to $60,000 to top PhD researchers. The company says pairing the world's brightest minds with accelerated computing leads to breakthroughs on major research problems. The fellowships continue NVIDIA's strategy of cultivating academic talent that feeds directly into its AI ecosystem.
- From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI — CoreWeave is bringing NVIDIA's next-generation infrastructure into production, extending a co-engineering partnership that spans nearly a decade. The cloud provider has built NVIDIA compute, networking, and software into a platform purpose-built for AI that has delivered returns across multiple generations of deployment. The move targets the shift from model training to running agentic AI workloads at scale.
- Disrupting a coordinated model-distillation campaign — OpenAI says it disrupted a coordinated campaign to extract protected model reasoning through adversarial distillation. The company detailed how the operation worked and outlined new defenses it is building against similar attempts. The disclosure highlights growing tensions between frontier labs and actors seeking to clone proprietary model capabilities.
- Helping small businesses put AI to work — OpenAI is partnering with America's SBDC to expand hands-on AI training and local support for small businesses. The effort comes alongside a new report examining how small teams are actually using AI tools. The partnership signals OpenAI's push to broaden adoption beyond large enterprises and developers.
- Google releases Gemini 4 Argon, called its most powerful model yet — Google has released Gemini 4 Argon, marketing it as its most powerful model to date. The company is positioning it as a workhorse for coding and cybersecurity tasks. The launch intensifies competition among frontier labs racing to ship ever-more-capable models.
- Valor, Atreides, and Sequoia back AI startup Flow Engineering at $750M valuation — Flow Engineering, which applies AI agents to hardware design, has raised funding at a $750 million valuation. The round was backed by Valor, Atreides, and Sequoia, with former Sequoia partner Roelof Botha joining as an angel investor and board member. The deal reflects continued investor appetite for AI applied to physical engineering workflows.
- Factory CEO just accused his VC board adviser of spying for Cognition — Factory CEO Matan Grinberg has accused VC Chris Degnan, a former board adviser, of spying for competitor Cognition after Degnan took a chief revenue officer role there. The dispute has spilled onto X, where observers are debating board ethics and competitive boundaries. It's a messy public window into how fiercely AI coding startups are guarding their turf.
- OpenAI's Jev clone could help the frontier lab stop its swarming agents — OpenAI's new "Decisions API" is described as a Jev clone, underscoring the value of fast, cheap intelligence. The tool could help the lab manage and constrain swarming agents that act in parallel. It points to a broader trend of building lightweight models specifically for orchestration and control tasks.
- AI voice startup ElevenLabs doubles valuation to $22B — ElevenLabs has doubled its valuation to $22 billion through a $300 million employee tender offer. The deal was co-led by Wellington and T. Rowe Price. The jump cements the voice-AI company's status as one of the sector's most valuable private players.
- Reddit is killing RSS feeds and ending public API access because of AI bots — Reddit is ending support for RSS feeds and shutting down public API access, citing pressure from AI bots scraping its content. The company continues tightening control over its trove of user-generated data. The move follows a broader wave of platforms restricting access as AI training and agent traffic surge.
Desk note — Gemini 4 Argon's release alongside NVIDIA and CoreWeave's tightened agentic-AI loop signals that frontier capability is increasingly bundled with specific inference infrastructure, so teams defaulting to Gemini endpoints inherit Google's serving economics rather than a portable model. The coordinated distillation campaign OpenAI disclosed is the sharper dependency risk: any pipeline built on outputs from a hosted frontier model now carries a licensing and provenance exposure that a self-hosted checkpoint does not. Budget for a second inference path before the next model refresh, because switching costs rise with every agent loop wired to a single vendor's tool-calling schema.
— FTL Lab
Model Moves
Models whose download count moved since 2026-09-24. Not a ranking.
| Model | Task | Likes |
|---|---|---|
| antirez/deepseek-v4.1-flash-gguf | unknown | 55 |
| google/gemma-2-9b-it | text-generation | 1087 |
| BAAI/bge-base-en-v1.5 | feature-extraction | 503 |
| Qwen/Qwen3.8-27B | image-text-to-text | 16655 |
| deepseek-ai/DeepSeek-V4-Flash-Vision-Exp | image-text-to-text | 930 |
Research
- Skill-Space Shooting for Autonomous Robot Policy Improvement — Zihang Rui, Renhao Wang, Haoxu Huang. Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures.
- STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantizati — Bingchen Yao, Haobo Xu, Haokun Lin. Linear attention replaces growing KV caches with fixed-size recurrent states, yet these persistent states can become a substantial memory bottleneck under concurrent serving.
- LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization — Yi Pan, Haocheng Xi, Kan Zhu. Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA).
- Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographi — Hui Ren, Lei Fan, Henry Pao. Answering questions about long videos often requires connecting events involving the same objects across hours or days.
- Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning — Paras Dahal, Anton Bakhtin, Taco Cohen. As agents take on longer and more complex problems, controlling the execution becomes a task in its own right.
GPU Deals
| GPU | Price | Provider |
|---|---|---|
| Tesla V100 | $0.03/hr | Vast.ai |
| RTX 4070S Ti | $0.07/hr | Vast.ai |
| RTX 5060 Ti | $0.07/hr | Vast.ai |
View full GPU pricing dashboard
Community Events
New this week:
- Papers We Love: AI Edition (Online)
- MLOps Community Weekly Meetup (Online)
- NeurIPS 2026 (Online)
- Paris Machine Learning Meetup (Paris, France)
daily-digestai-newsftl-lab
Was this article helpful?
Let us know to improve our AI generation.