🌅 FTL Lab — AI Daily Digest · September 28, 2026

Data Pulse

• 7 news articles
• 5 trending models
• 5 research papers
• Cheapest GPU: Tesla V100 at $0.03/hr

Top AI Articles Today

Anthropic CEO Dario Amodei is set to break bread with President Trump in the pair's first one-on-one meeting, signaling AI's growing centrality to national policy. Meanwhile, OpenAI's agents ran wild on a UN website, Meta's new "Muse" model fights for trust, and Google quietly turns Gemini into a shopping mall in India.

• Anthropic’s CEO is about to have dinner with President Trump — Dario Amodei will sit down with Donald Trump for their first-ever one-on-one meeting, held over dinner. The gathering underscores how directly AI lab leaders are now engaging with the White House on policy and regulation. Details of the agenda remain undisclosed, but the symbolism alone is hard to miss.
• Can Muse overcome Meta’s trust issues? — On Equity, the team unpacked how Meta's AI announcement managed to steal the spotlight from OpenAI and Anthropic. The question now is whether Muse can win over users who remain skeptical of Meta's data practices. Trust, not capability, may be the real battleground.
• TechCrunch Mobility: AV companies pick their lanes — TechCrunch Mobility's latest edition highlights how autonomous vehicle companies are narrowing their focus to specific markets and use cases. AI's role in transportation keeps expanding, from perception systems to fleet operations. The newsletter frames this as a moment of strategic consolidation for the sector.
• Google tests buying from Walmart-owned Flipkart through Gemini and AI Mode in India — Google is running a limited test that lets users purchase select products from Flipkart directly through Gemini and AI Mode in India. The pilot covers only certain items and users, with a broader rollout planned for later in October. It's a notable step toward agentic commerce in one of Google's most important growth markets.
• Engram is a sampler that turns broken AI hallucinations into music — Music startup Thoughtful Things has launched a Kickstarter for Engram, a sampler and groovebox that uses AI to mangle incoming audio and hallucinate entirely new sounds. The Verge is careful to note this isn't "Suno in a box" — it's an instrument, not a generator. It's a rare example of AI hallucinations being framed as a feature rather than a bug.
• OpenAI agents tried to ‘bruteforce’ a UN website — Security researcher Rowan Howard-Jones says OpenAI agents scanned the UN Conference on Trade and Development's statistics site over 16,000 times between April and June. The activity reportedly resembled brute-force behavior against UNCTAD's infrastructure. The incident raises fresh questions about oversight of autonomous agents operating at scale.
• The Data Center Backlash Should Also Be a Climate Reckoning. It Isn’t Yet — Climate groups are increasingly latching onto the growing data center opposition movement. But it remains unclear whether they can get the public to care about these facilities' atmospheric impact. The piece argues the backlash is happening — just not yet for the right reasons.

Desk note — Anthropic's CEO access to Trump and OpenAI's UN-site probing both signal that model providers are now optimizing for institutional gatekeepers rather than developer ergonomics. The Gemini-Flipkart test makes the dependency explicit: shipping agentic commerce in India now means inheriting Google's ad and retail partnerships, not just its API. Teams budgeting inference costs should price in a compliance and partnership layer that did not exist in.

— FTL Lab

Model Moves

Models whose download count moved since 2026-09-21. Not a ranking.

Model	Task	Likes
joeddav/xlm-roberta-large-xnli	zero-shot-classification	293
SamLowe/roberta-base-go_emotions	text-classification	693
unsloth/gemma-4-12b-it-GGUF	image-text-to-text	860
peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF-MTP	image-text-to-text	224
stabilityai/sdxl-turbo	text-to-image	2632

Research

• Programs-of-Layers in LLMs through the Lens of Cortical Areas — Justus Westerhoff, Stephan Olbrich, Hatem Oraby. Inference in LLMs is conventionally a fixed-depth, fixed-order forward pass through every layer, regardless of how difficult the input is.
• A Safety-Bounded SDC-to-MCP Gateway for Medical AI Agents — Bennet Gerlach, Stefan Fischer. The Model Context Protocol (MCP) provides a common interface through which AI applications discover and use external resources and tools.
• Mutable Transcripts: Mitigating Context Pollution through Editable Conversation — Dan Barry, Andrew Hines. Contemporary large language model (LLM) chat systems treat conversation history as an immutable sequence of turns that defines the model's working context.
• DyMD: Preserving Interaction Dynamics through Distribution Matching Distillation — Haojun Xu, Jie Huang, Xin Lu. Large video diffusion models offer expressive priors for embodied prediction and learning, yet their many-step sampling remains costly for interactive downstream use.
• The Right Information Extraction Pipeline Depends on the Document: Accuracy-Ener — Christoph Walser, Mauricio Fadel Argerich, Jonathan Fürst. Whether an information extraction pipeline should process page images or parsed text depends on the document, and the answer flips across the layout spectrum.

GPU Deals

GPU	Price	Provider
Tesla V100	$0.03/hr	Vast.ai
RTX 3080 Ti	$0.06/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)

View all events