All entries

Watch the goal post shift unfold in real time: AGI used to be doing anything a person, including an expert, could do, and by the end of the interview it’s not. I call this the AI bait and switch. As I

We are excited to join Nvidia's Nemotron Coalition of leading AI labs working together to advance open frontier foundation models. To celebrate we have partnered with @nvidia and @nebiustf to provide

We are saying no to a data centre in Île-des-Chênes because there are big threats to the environment and not much benefit to the economy. Our message to any tech company out there… if you want to buil

We built a cool project that shows you how to compose our core document intelligence primitives into a reusable pipeline that turn your documents into machine-readable data. Come check it out! https:/

We have a Fleet agent called @docs_plz in our Slack that's made a very noticeable impact on our velocity of docs changes. In the chart below, you can clearly see that after it was added, the amount of

We partnered with Shopify so you can go from idea to live store in minutes Just tell Replit Agent what you want to sell. It will: - Build a custom storefront - Create your Shopify store - Help you add

We spent the week at #MSBuild talking about one thing: fine-tuning has gone from 'maybe not worth it' to your actual competitive moat. @lqiao sat down with @yina_arenas to break down why and what Fire

arXiv:2606.04060v1 Announce Type: new Abstract: Weakly Incremental Learning for Semantic Segmentation (WILSS) suffers from the continuous introduction of noisy supervision, which progressively corrupt

We're building in Canada. 🇨🇦 For decades, Canada invested to build the research foundations that made modern AI possible. Now we have to build, train, and scale what comes next here at home. Canada's

We're kicking off our first cohort of Runway Builders with almost 100 startups building at the frontier of real time video with the Runway API – notifications going out to accepted companies today! If

We're presenting ParseBench at CVPR 2026! ParseBench is the most comprehensive document understanding benchmark for VLMs. ✅ It contains 2k pages of real-world enterprise documents ✅ It has comprehensi

We're presenting ParseBench at CVPR 2026 today. 🦙 Come learn why document understanding is an AGI-complete problem (an agent can't act on a doc it can't correctly read, and reading a real enterprise t

we're taking an early bet on open models specifically, because they're SO much cheaper one point of reference: an app outputting 10M tokens/day costs roughly 250/day on Opus 4.8 versus ~24/day for Min

arXiv:2507.03373v2 Announce Type: replace Abstract: Given Wikipedia's role as a trusted source of high-quality, reliable content, concerns are growing about the proliferation of low-quality machine-ge

We’ve been researching new ways for ChatGPT memory to carry context across conversations and keep it useful over time. Today, that work is rolling out as a more capable memory system in ChatGPT. https

arXiv:2606.04233v1 Announce Type: new Abstract: A robotics benchmark score measures success under one fixed evaluation setup, yet is routinely treated as evidence of general manipulation capability. W

arXiv:2606.04989v1 Announce Type: cross Abstract: Although much is known about the physical danger of cycling situations, less is understood about the perceived danger of cycling. Furthermore, percept

What happened when one of our models found a counterexample to an 80-year-old Erdős conjecture? Researchers @alexwei_, @HongxunWu, and @wjmzbmr1 shared the story on the OpenAI Podcast with @AndrewMayn

arXiv:2606.04425v1 Announce Type: cross Abstract: Modern agentic systems transform LLMs from session-bounded assistants into stateful systems that persist and evolve shared world state across sessions

what makes building on replit different? it all happens in one place. → describe your idea in plain english, get working software → generate UI, add auth + databases, deploy → collaborate with your te

Enterprise AI is forcing companies to rethink where their most sensitive workloads should live, and private cloud is moving deeper into the center of that decision. For many organizations, the cloud c

arXiv:2606.04935v1 Announce Type: new Abstract: Active inference casts decision-making as inference, with the Expected Free Energy (EFE) unifying goal-directed and information-seeking behavior. Recent

At Google Cloud, our goal is to let you run large-scale analytical and data science workloads with maximum efficiency so you can process big data pipelines, machine learning, and ETL tasks. We recentl

arXiv:2502.08870v2 Announce Type: replace Abstract: We provide an approach for the analysis of randomised exploration algorithms like Thompson sampling that does not rely on forced optimism or posteri

arXiv:2606.04168v1 Announce Type: new Abstract: Safety alignment in large language models (LLMs) is fragile in part because it is often shallow: fine-tuning mainly reshapes the model's behavior near t

arXiv:2606.04476v1 Announce Type: new Abstract: In this paper, we study the gradient descent dynamics for jointly training both layers of a one-hidden-layer ReLU network to fit a linear target functio

arXiv:2606.04389v1 Announce Type: new Abstract: Large Language Models (LLMs) show promise in psychological counseling, yet existing benchmarks rely heavily on highly cooperative simulated clients. We

arXiv:2603.09242v2 Announce Type: replace Abstract: The growing realism of generative models has blurred the boundary between real and synthetic content, posing significant challenges to reliable AI-g

arXiv:2606.04375v1 Announce Type: new Abstract: Differentially private stochastic gradient descent (DP-SGD) injects noise into every updated coordinate, making the injected noise energy scale with the

arXiv:2606.04361v1 Announce Type: cross Abstract: Age of Information (AoI) has become a central metric for the design of wireless update systems, especially in applications where fresh measurements su

arXiv:2606.04161v1 Announce Type: new Abstract: Different predictors often excel on different inputs, so picking the best one per instance promises higher accuracy than committing to a single model. I

arXiv:2606.04127v1 Announce Type: new Abstract: Medical question answering is a high-stakes setting where factual errors can have serious consequences. Retrieval-augmented generation (RAG) is widely v

arXiv:2606.04098v1 Announce Type: new Abstract: Video misinformation increasingly operates at the semantic and evidential level: authentic footage may be selectively edited, temporally reordered, spli

When you burn so much money you run out of options… Anthropic co-founder and President Daniela Amodei said the high cost of developing AI models is driving firms like hers to look to the public market

Gemma 4 12B is the first medium-sized, encoder-free multimodal model capable of natively ingesting audio and video , recently released by Google. The model is available on Ollama with 11.7M downloads

arXiv:2606.05107v1 Announce Type: cross Abstract: We propose a label-free approach to adapt powerful but generic vision foundation models to specialized scientific domains. Standard supervised fine-tu

⚠️Why didn’t the hyperscalers wait until after the IPO in switching to pay-by-usage-charging? My guess is that *they literally could not afford to* —because it would bankrupt them. My reasoning: in “a

Why Japanese companies do so many different things: The internal logic of the world’s strangest corporations https://davidoks.blog/p/why-japanese-companies-do-so-many TL;DR > you have a firm that has

arXiv:2606.04662v1 Announce Type: cross Abstract: Muon improves training efficiency over Adam in large language-model training by about two times, but the local geometric source of this advantage rema

Agentic AI is making the semantic layer an essential enterprise priority because headless agents asking thousands of questions simultaneously have zero tolerance for inconsistent data definitions. Thi

Cursor has launched a feature enabling users to publish and share canvases—custom applications such as dashboards, reports, and internal tools—with their teams. This update extends Cursor's canvas fun

OpenAI introduced a new memory system for ChatGPT that allows users to review and control what the AI remembers across conversations through a memory summary feature. This update provides enhanced tra

This post appears to be a query from Swyx asking @_catwu for an update on a previously shared chart, specifically regarding developments or changes that occurred after the release of Opus 4.8. The pos

This entry documents a Wordle game result where the player solved puzzle #1,810 in 4 attempts, using the color-coded feedback system (black for incorrect letters, yellow for correct letters in wrong p

arXiv:2606.04916v1 Announce Type: new Abstract: Worker utility is not observed -- only its consequence is. Each gig transaction produces a single bit: accepted or rejected. We argue this structure poi

arXiv:2606.05159v1 Announce Type: new Abstract: Rigorous evaluation of learning-based robotic systems is an essential prerequisite for deployment. However, real-world test data is expensive to gather;