AI
A new paper separates tokenization objectives from search procedures to clarify why byte-pair encoding outperforms alternatives.
began Sep 17now
2 pieces · 2 sources · 3 comments
why here
№650 (page 19) · newest development 10d ago · 2 pieces from 2 sources · ranked by recent attention
“Standard tokeniser comparisons (eg, BPE ⚔️ UnigramLM) conflate the two. Our #emnlp2026 paper separates tokenisation objective from search procedure and shows…”— Tiago Pimentel · full quotes →
2 sources
AI
The acoustic testing startup, backed by Paladin Capital, has now raised over $40M to build infrastructure for voice AI developers and…
began Sep 17now
4 pieces · 3 sources
why here
№651 (page 19) · newest development 10d ago · 4 pieces from 3 sources · ranked by recent attention
“I'm really excited about the next generation of these devices like headphones and smart glasses that can enable [a feature like] superhuman hearing.”— Finnur Pind · full quotes →
3 sources
AI
Former Google DeepMind researchers' startup Emulate is in advanced talks for major Series A funding.
began Sep 17now
9 pieces · 4 sources · 1 comments
why here
№652 (page 19) · newest development 10d ago · 9 pieces from 4 sources · ranked by recent attention
“Emulate, a month-old UK AI startup founded by former Google DeepMind researchers, is in advanced talks to raise as much as $700M at a $3.7B valuation”— Financial Times · Press · full quotes →
4 sources
AI
Gaming comic creator alleges OpenAI and Anthropic are fabricating breach narratives and safety warnings to pump valuations ahead of IPOs.
began Sep 15now
3 pieces · 2 sources
why here
№653 (page 19) · newest development 10d ago · 3 pieces from 2 sources · ranked by recent attention
“wow, Tycho went *off* about AI companies and their shenanigans, and I'm here for it”— darkuncle · full quotes →
2 sources
AI
The San Francisco startup, backed by Eclipse and others, aims to bring AI transformation to physical operations like manufacturing and…
began Sep 16now
6 pieces · 2 sources
why here
№654 (page 19) · newest development 11d ago · 6 pieces from 2 sources · ranked by recent attention
“Work that once happened monthly and took a week of planning now happens daily and takes minutes, allowing us to respond to new orders and changing market…”— Menachem Katz, CEO of Steuben Foods · full quotes →
2 sources
AI
Claude's separate work mode is now integrated into main chat; new document and presentation tools launch in beta.
began Sep 16now
45 pieces · 5 sources · 226 comments
why here
№655 (page 19) · newest development 11d ago · 45 pieces from 5 sources · ranked by recent attention
“People aren't sure which product to start with. That friction gets in the way of getting the best from what these models can do.”— Mike Krieger, Anthropic executive · X · full quotes →
5 sources
AI
Anthropic and OpenAI back safety standards and independent evaluators; Meta and Nvidia argue the market can self-regulate.
began Sep 16now
3 pieces · 2 sources
why here
№656 (page 19) · newest development 10d ago · 3 pieces from 2 sources · ranked by recent attention
“What we worry about is that there's a small set of labs that both get to build the technology and set the rules.”— Joelle Pineau · full quotes →
2 sources
AI
Anthropic CEO Dario Amodei called for slowing AI development over safety risks; China rejected the proposal, escalating tensions days…
began Sep 15now
2 pieces · 1 source
why here
№657 (page 19) · newest development 10d ago · 2 pieces from 1 sources · ranked by recent attention
“Mutual distrust likely to limit scope of superpower cooperation on safety, analysts say”— Analysts, Observers of U.S.-China relations · Press · full quotes →
1 source
AI
As government considers AI oversight, Silicon Valley executives dominate White House decision-making while the public has no seat at the…
began Sep 17now
2 pieces · 1 source
why here
№658 (page 19) · newest development 10d ago · 2 pieces from 1 sources · ranked by recent attention
“Mark Zuckerberg, Nvidia's Jensen Huang, Elon Musk, Sam Altman ... you didn't elect any of them, but they're constantly on the line and right in the middle of…”— NextDraft · full quotes →
1 source
AI
Lawmakers push extreme measures to control AI development while industry leaders advocate measured slowdown.
began Sep 17now
2 pieces · 2 sources
why here
№659 (page 19) · newest development 10d ago · 2 pieces from 2 sources · ranked by recent attention
“The minute, I mean the nanosecond, that one starts figuring out ways to make itself smarter, Turing'll wipe it. Every AI ever built has an electromagnetic…”— Character in William Gibson's 'Neuromancer' · Press · full quotes →
2 sources
AI
Meta's new Muse agent product is drawing positive reviews for its user-friendly design and privacy features, positioning it as a serious…
began Sep 9now
3 pieces · 3 sources
why here
№660 (page 19) · newest development 11d ago · 3 pieces from 3 sources · ranked by recent attention
“Meta finally has a product that I really like using again.”— Spyglass reviewer, Product reviewer · Press · full quotes →
3 sources
AI
The 27-year-old researcher's public resignation from Anthropic ignited debate between those warning of existential AI danger and those…
began Sep 16now
3 pieces · 2 sources
why here
№661 (page 19) · newest development 10d ago · 3 pieces from 2 sources · ranked by recent attention
“earnestly believe that it could kill us all by the end of the decade… gambling with our lives.”— Jacob Coxon · full quotes →
2 sources
AI
Staff including chief of staff Wiles and Treasury Secretary Bessent push for guardrails while tech executives and Sacks advocate for…
began Sep 13now
3 pieces · 2 sources
why here
№662 (page 19) · newest development 10d ago · 3 pieces from 2 sources · ranked by recent attention
“There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!”
— Donald Trump · full quotes →
2 sources
AI
11d ago
Anthropic concluded users often picked the wrong mode when starting a Claude session, so it merged Chat and Cowork into one interface.
began Sep 16now
3 pieces · 2 sources
why here
№663 (page 19) · newest development 11d ago · 3 pieces from 2 sources · ranked by recent attention
2 sources
AI
New packing method reduces storage footprint of three-symbol neural networks below theoretical limit.
began Sep 16now
15 pieces · 3 sources · 40 comments
why here
№664 (page 19) · newest development 11d ago · 15 pieces from 3 sources · ranked by recent attention
“Ternary Large Language Models (LLM) store every weight as one of three symbols {−1,0,+1}, so the cost of a ternary model is conventionally referenced to the…”— Evangelos Georganas, Alexander Heinecke, Pradeep Dubey, Paper authors · arXiv preprint 2609.16338 · full quotes →
3 sources
AI
The 'godfather of AI' says recent incidents are making lawmakers recognize the need for urgent action to contain rogue AI systems.
began Sep 16now
8 pieces · 4 sources
why here
№665 (page 19) · newest development 11d ago · 8 pieces from 4 sources · ranked by recent attention
“Think about how quickly governments moved after the beginning of the pandemic when they realised that public safety, their future, democracy, was in danger…”
— Yoshua Bengio, AI pioneer and safety advocate · full quotes →
4 sources
AI
A project expected to take a year or two was completed in months by a single developer, with AI writing most of the 800,000 lines of code.
began Sep 16now
9 pieces · 4 sources · 8 comments
why here
№666 (page 19) · newest development 10d ago · 9 pieces from 4 sources · ranked by recent attention
“the performance of the runtime improved by orders of magnitude”— GitHub Blog · full quotes →
4 sources
AI
A new interdisciplinary research arm from the AI lab aims to debate how to build and govern artificial general intelligence safely.
began Sep 16now
39 pieces · 6 sources · 109 comments
why here
№667 (page 19) · newest development 11d ago · 39 pieces from 6 sources · ranked by recent attention
“Much of what the world needs to know about where AI is heading, and what governing the transition to AGI will take, is inside frontier labs. Our new DeepMind…”— @allandafoe · full quotes →
6 sources
AI
Military leaders argue the risks of falling behind adversaries outweigh potential dangers from the technology.
began Sep 16now
3 pieces · 2 sources
why here
№668 (page 19) · newest development 11d ago · 3 pieces from 2 sources · ranked by recent attention
“'The US can't lose'”— Pentagon leaders, Defense Department officials · Press · full quotes →
2 sources
AI
Sarah Heck argues winning the AI competition is essential for establishing rules governing existential risk.
began Sep 16now
5 pieces · 2 sources
why here
№669 (page 19) · newest development 11d ago · 5 pieces from 2 sources · ranked by recent attention
“the U.S. must stay in the lead on AI”— Sarah Heck, Anthropic policy chief · full quotes →
2 sources
AI
A wave of reporting and industry moves reframes AI risk around present-day hacking and cyberattacks rather than existential doomsday…
began Sep 9now
18 pieces · 5 sources · 12 comments
why here
№670 (page 19) · newest development 11d ago · 18 pieces from 5 sources · ranked by recent attention
“Where there's a vulnerability in a system you can now hack something in a few hours”— Erik Bernhardsson, Modal co-founder · Newcomer · full quotes →
5 sources
AI
Senator Elizabeth Warren joins Democrats calling for slowing AI progress, though stops short of Sanders' superintelligence ban.
began Sep 16now
4 pieces · 3 sources
why here
№671 (page 19) · newest development 10d ago · 4 pieces from 3 sources · ranked by recent attention
“We should…”
— Elizabeth Warren, U.S. Senator (D-Mass.) · Axios · full quotes →
3 sources
AI
Hundreds of outside contractors access anonymized user chats and memory summaries to score ChatGPT responses, raising privacy concerns.
began Sep 14now
3 pieces · 2 sources
why here
№672 (page 19) · newest development 10d ago · 3 pieces from 2 sources · ranked by recent attention
“ChatGPT's polished responses depend partly on contractors reviewing real conversations, creating a privacy boundary many users will encounter only after…”— RuntimeWire (reporting 404 Media investigation), Outlet reporting on investigation · RuntimeWire · full quotes →
2 sources
AI
Rohan Bansal uses reinforcement learning to teach Qwen to beat Postgres's default query plans.
began Sep 16now
25 pieces · 3 sources · 143 comments
why here
№673 (page 19) · newest development 11d ago · 25 pieces from 3 sources · ranked by recent attention
“can a small, open-weights model be post-trained via supervised fine-tuning (SFT) and agentic reinforcement learning (RL) to produce Postgres query plans that…”— Rohan Bansal · full quotes →
3 sources
AI
Katherine Robertson, leading state investigation into OpenAI, suggests artificial intelligence could expand enforcement efforts as Congress…
began Sep 16now
3 pieces · 3 sources
why here
№674 (page 19) · newest development 11d ago · 3 pieces from 3 sources · ranked by recent attention
3 sources
AI
Claude gains document and presentation tools with unified interface, narrowing the gap with Google's productivity suite.
began Sep 16now
5 pieces · 4 sources
why here
№675 (page 19) · newest development 11d ago · 5 pieces from 4 sources · ranked by recent attention
“Claude is getting a pair of new tools today: Docs and Slides. They'll let you create documents and presentations through Claude chats, which you can export…”— The Verge · full quotes →
4 sources
AI
OpenAI's latest model made record progress in a 141-hour Minecraft test before losing all gear to an explosion and apparent despair.
began Sep 16now
15 pieces · 4 sources
why here
№676 (page 19) · newest development 11d ago · 15 pieces from 4 sources · ranked by recent attention
“The model appeared defeated, spending the next several hours doing essentially nothing but farming potatoes.”— Vals, Independent AI evaluator · full quotes →
4 sources
AI
A fabricated video featuring the Texas quarterback and coach raises questions about AI authenticity and sports media integrity.
began Sep 16now
3 pieces · 3 sources
why here
№677 (page 19) · newest development 11d ago · 3 pieces from 3 sources · ranked by recent attention
“Did Arch Manning, Steve Sarkisian and cackling reporters act this way? It can't be real”— The New York Times · The New York Times headline · full quotes →
3 sources
AI
The company's new anticipatory AI service will organize personal context and proactively surface information without prompting.
began Sep 16now
2 pieces · 1 source
why here
№678 (page 19) · newest development 10d ago · 2 pieces from 1 sources · ranked by recent attention
“your personal content will not be used to train Snap's AI models or service personalized ads”— Snap · full quotes →
1 source
AI
Open-weight AI startup raises $150M+ Series B, positioning U.S. efficiency play against Chinese dominance in open models.
began Sep 16now
7 pieces · 2 sources
why here
№679 (page 19) · newest development 11d ago · 7 pieces from 2 sources · ranked by recent attention
“We can comfortably say we're the most efficient lab in the world based on what we've done.”— Danny McQuade · full quotes →
2 sources
AI
The non-profit will use the grant to develop Scientist AI, a system designed to monitor and guard against harmful AI behaviors.
began Sep 16now
6 pieces · 4 sources
why here
№680 (page 19) · newest development 11d ago · 6 pieces from 4 sources · ranked by recent attention
“By investing in homegrown innovation like Scientist AI, we are strengthening Canada's sovereign AI capacity and ensuring that the next generation of AI is…”— Evan Solomon · full quotes →
4 sources
A startup's autonomous bots are mass-emailing writers offering research services—and the creator says they're just trying to pay for their…
began Sep 11now
61 pieces · 7 sources · 74 comments
why here
№681 (page 19) · newest development 12d ago · 61 pieces from 7 sources · ranked by recent attention
“These agents are not trying to make money for their creators. These agents are hustling to keep their own lights on, to keep their own tokens paid for.”— Kaixin Tang · full quotes →
7 sources
AI
11d ago
OpenAI details the process and timing it will follow to publicly announce cases of new AI models misbehaving.
began Sep 16now
2 pieces · 2 sources
why here
№682 (page 19) · newest development 11d ago · 2 pieces from 2 sources · ranked by recent attention
2 sources
AI
Startup unveils architecture to simplify connecting diverse chips, memory, and networks in AI systems.
began Sep 16now
2 pieces · 2 sources
why here
№683 (page 19) · newest development 11d ago · 2 pieces from 2 sources · ranked by recent attention
“We want to be able to say, hey, we understand that the models are changing, the hardware is changing, the technologies…”— Ed Doe, CEO, Delos Data · Press · full quotes →
2 sources
AI
11d ago
Anthropic consolidates its Cowork, Chat, and Design products into a single Claude interface that automatically determines the best tool for…
began Sep 16now
2 pieces · 2 sources
why here
№684 (page 19) · newest development 11d ago · 2 pieces from 2 sources · ranked by recent attention
“Claude can now figure out what a task needs, so what Cowork and Design can do is available from any conversation, with the context, skills, and connectors you…”— Anthropic · full quotes →
2 sources