标签：#AI

Decart is launching Oasis 3, a real-time world model that generates photorealistic driving environments for autonomous vehicle testing, now available via API for developers to build on.

Insta360 has launched its first gimbal camera, the $770 Luna Ultra.

The generation side is shipping fast (TileGym, AutoKernel, KernelEvolve). The validation-side surface for “what the kernel actually did at runtime” has not kept pace. TL;DR In the past nine months, three significant releases have landed for auto-generation of CUDA kernels: NVIDIA TileGym , RightNow AutoKernel, and Meta’s KernelEvolve. Each ships training infrastructure for kernel generation. Validation infrastructure (what the generated kernel actually did at runtime, on a real workload, in a production-shaped environment) has not kept the same pace. eBPF traces are the ground-truth layer that closes the gap. What “validation” means at the kernel level Two distinct validation surfaces: Pre-launch: the generated CUDA C compiles, the PTX assembles, the kernel passes a numerical-equivalence test against a reference. Standard compiler / unit-test territory. Generation frameworks ship this themselves. Post-launch: the kernel ran, returned, took N microseconds, used M registers per thread, hit X cache miss rate, and did or did not serialize the rest of the stream behind it. This is the layer that an eBPF trace plus standard CUDA driver counters can answer for any kernel, generated or hand-written. Auto-generation pipelines do not by default close the post-launch loop. They demonstrate “the kernel works in our test setup”. They do not demonstrate “the kernel does not regress p99 latency on production inference traffic”. What an eBPF trace adds to a generated kernel Once a generated kernel is in a real workload, the same trace surface used for any CUDA kernel applies: launch latency from cudaLaunchKernel , sync stalls from cudaStreamSynchronize , host-side overhead from the dispatcher, host scheduling preemption while the GPU is busy. None of those signals are visible to a generation framework that evaluates kernels in isolation. -- post-launch validation: did the new generated kernel regress p99? SELECT kernel_name , COUNT ( * ) AS launches , AVG ( duration_ns ) / 1 e3 AS

Anthropic's strongest model is free until June 22 — and two more shifts for builders Three things landed for builders at once: the best model got cheaper (free, actually), free inference showed up on Apple's stack, and one still photo now becomes a talking video. Two of them you can act on right now. Here's the 90-second video version if you want the quick pass first: 1. Claude Fable 5 is public — and free on your plan until June 22 Anthropic released Claude Fable 5 , the first publicly available version of its Mythos-class model. It's state-of-the-art on nearly every benchmark Anthropic tests — software engineering, knowledge work, vision, and scientific research. It's free on Pro, Max, Team, and Enterprise plans through June 22 ; after that it's 10 dollars per million input tokens and 50 per million output . In high-risk areas (cyber, bio, chem) it refuses and falls back to Claude Opus 4.8 — about 95% of Fable sessions run entirely on Fable. This dropped just days after Anthropic publicly warned that AI was getting too dangerous. Why it matters: the strongest Claude is free to try on your existing plan for a two-week window. Run your hardest real task on it now and benchmark it before June 22 — the kind of jump that's worth re-checking your evals against. 2. Apple made its Foundation Models free for small developers At WWDC 2026 , Apple gave developers in the App Store Small Business Program (apps under 2 million first-time downloads ) free access to the next generation of Apple Foundation Models running on Private Cloud Compute — removing inference cost as a barrier. The Foundation Models framework now supports image input . A single Swift API can also call third-party models like Claude and Gemini, server-side. A new Dynamic Profiles system supports multi-agent workflows, and Apple will open-source the framework later this summer. Why it matters: you can ship AI features into an app without an inference bill. Prototype on Apple's free on-device models, and route

Uber's CTO told the world this month that the company spent its entire 2026 AI allocation by April. The story has been reported in a handful of outlets, hit the front page of Hacker News for 397 points and 469 comments , and is mostly being read as a cost-of-AI-tools story. It is one. It is also, on a closer reading of the numbers, a pricing-model story — and the structural fact that almost none of the coverage has emphasized is the one that determines whether this is a one-company anomaly or the beginning of an industry-wide budgetary crisis. The structural fact is that Claude Code, like most enterprise AI tooling in 2026, is priced on token consumption, not per-seat licensing. Token-based pricing scales with how aggressively the tool is used. Per-seat enterprise SaaS pricing — the model corporate IT budgets are built around — scales with how many people have access to it. Those two cost curves diverge in exactly the territory where productivity tools are designed to operate: high-engagement, daily-use, gradually-deepening workflows. The Uber data is the first public-facing version of a math problem most enterprise IT departments are about to discover privately. The numbers Uber CTO Praveen Neppalli Naga , named in Yahoo Finance's and Benzinga's coverage, said publicly that Uber is "back to the drawing board" on AI budgeting after the surge in Claude Code use blew through internal projections. The specific numbers, as reported across the multiple outlets covering the story: Claude Code adoption inside Uber's ~5,000-engineer organization went from 32% to 84% over four months. 70% of committed code at Uber is now AI-originated. 11% of live backend updates are "being written by AI agents built primarily with Claude Code," per the reporting. Per-engineer monthly API costs: $500 to $2,000. Uber's annual R&D spend is around $3.4 billion , of which the AI tooling line was a much larger fraction than expected. Cursor adoption plateaued; Claude Code dominated. These are ext

We train a model on handwritten digit classification. 99% accuracy . Then we train the same model on a new task — say, fashion item recognition. We go back and test it on digits. 34% accuracy . It has completely forgotten. Not gradually, not partially — almost entirely. What Just Happened? We trained a CNN on MNIST digits — 99.2% accuracy . After fine‑tuning on Fashion MNIST, it reached 91.1% accuracy . But when re‑evaluated on MNIST, accuracy collapsed to 33.9% . This collapse is catastrophic forgetting : the model’s weights shifted to optimize for the new task, erasing the old solution. Why did training on more data make the model worse at something it already knew? MNIST is handwritten digits (0–9). Fashion MNIST is clothing items like shirts and shoes. Both are 28×28 grayscale images, but the tasks are distinct. Why Does It Happen? The core issue is that the model relies on the same set of weights for both tasks. There is no separation or dedicated memory; every parameter is shared . When training shifts from Task A ( MNIST digits ) to Task B ( Fashion MNIST ), gradient descent simply minimizes the loss on the data it sees at that moment. It has no awareness that Task A ever existed. In the loss landscape, imagine two parabolic bowls: one for Task A and one for Task B. The optimum for Task A lies at θ A ∗ ​ , while Task B's optimum is at θ B ∗ ​ . As training on Task B progresses, the weights θ move towards θ B ∗ ​ . This movement inevitably raises the loss for Task A because its minimum is left behind. The root cause is the shared weight space. Gradient descent is a stateless optimizer; it only follows the current gradient signal. Since the minima for Task A and Task B are far apart, there is no single configuration of θ that satisfies both tasks simultaneously. This is why catastrophic forgetting occurs. Weight space can be visualized as an N-dimensional space, where each axis corresponds to one parameter. Every point in this space represents a full set of wei

OAuth for Remote MCP Servers How each AI assistant signs in to a remote MCP (Model Context Protocol) server, and why the flow differs by client and by where it runs. Overview The protocol throughout is standard OAuth 2.1 — an open, widely implemented authorization standard. The human sign-in runs through oauth2-proxy , one of the most widely deployed open-source auth proxies; the only deployment-specific piece is a thin, spec-conforming authorization server (the /oauth endpoints) that hands MCP clients their tokens. Every client ends up the same way — a person signs in against Google (restricted to your organization's domain), and the client holds a short-lived bearer token it presents on each /mcp call. Two things differ between assistants: where the client runs (a machine on the VPN — private — vs. the vendor's cloud — public ), which decides the host it reaches; and what kind of OAuth client it is — a public client proving itself with PKCE (Proof Key for Code Exchange, which lets a client with no secret prove the token request comes from the same client that started the flow), or a confidential client proving itself with a secret. The participants oauth2-proxy — the public-facing reverse proxy. It authenticates the human against Google (the sign-in restricted to your organization's domain) and forwards the verified identity to the app behind it. Only oauth2-proxy faces the internet. It is a mature, heavily-deployed open-source project — the standard way to put Google/OIDC (OpenID Connect) single sign-on in front of a service, widely used in Kubernetes deployments — so the most security-sensitive leg of the flow (the OAuth exchange with the identity provider) runs on battle-tested code. The MCP server — the app on a loopback port behind the proxy. It plays two roles: the OAuth authorization server ( /oauth/authorize , /oauth/token , /oauth/register , .well-known discovery) and the /mcp tool endpoint. It mints codes and tokens, and validates a token on every /mcp c

I spent a weekend wiring Google's Gemini and Veo APIs into a single app just to feel where the edges of multimodal AI actually are. It turned into a small studio I now use daily, and along the way I learned more about these models from plumbing them than from any paper. Here's the honest technical debrief. Three pipelines, three completely different problems I wanted one prompt box that could do video, image editing, and document Q&A. Naively I assumed they'd share most of the stack. They don't. 1. Image-to-video: the enemy is time, not pixels Generating one good frame is solved. Video is about temporal coherence — frame 13 must agree with frame 12 or you get flicker and identity drift. Modern video models treat the clip as one object in space and time (latent diffusion over a width x height x time volume, with spatiotemporal attention) rather than 120 independent images. Conditioning on a reference image as the first frame is what makes image-to-video feel controlled: you've handed the model a strong anchor and asked it to extrapolate motion, not invent a world. The surprise: native audio sync (Veo 3.1 generating clip + soundtrack jointly) does more for perceived realism than another notch of resolution. A door slam landing on the exact frame the door shuts is uncanny in a good way. 2. Instruction-based image editing: preservation is the hard part Generating is unconstrained; editing must change one thing and preserve everything else. Condition the diffusion model on both the instruction and the source image's latents, cross-attend the instruction to steer only the referenced region, and bias hard toward preserving unedited latents. Push that preservation too soft and the subject's face quietly morphs across edits — the classic 'character consistency' failure that makes or breaks storytelling use-cases. 3. PDF chat: it's retrieval, not a long context The naive 'paste the whole PDF' approach dies on long files (models get lost in the middle ) and costs you the full

What device do we use almost all the time? Our mobile phone, almost certainly. If we wanted to...

Adi Polak discusses the architecture required to transition from stateless prompts to state-aware, context-rich AI agents. Drawing on 15 years in distributed systems, she shares how engineering leaders can leverage Apache Kafka and Flink for real-time stream processing, dynamic memory tiering, and tool orchestration via MCP to solve token limits, cost spikes, and latency bottlenecks. By Adi Polak

A new report from OpenAI details PRC-linked influence operations using AI to target U.S. tech debates, data center narratives, tariffs, and false claims about ChatGPT.

Ambrosia Energy wants to build power plants in less than 12 months while undercutting natural gas. It hopes to build gigawatts worth by 2030.

Doctor Who's previously-announced Christmas special is no more as the BBC severs ties with Russell T. Davies and Bad Wolf.

Over the last few weeks, while learning LangGraph and agentic systems, I ended up building Co-Founder Memory . It's a stateful AI assistant with: • long-term memory • planning loops • self-correcting RAG • web search fallback • automated timeline summaries • project and preference tracking Nothing revolutionary — many ideas already exist. The goal wasn't to reinvent memory, but to understand how these systems work by actually building one. A lot of concepts only started making sense once I had to connect them together: graph-based workflows with LangGraph memory extraction and storage retrieval and validation loops routing and planning nodes maintaining context across sessions Building it taught me far more than watching tutorials ever did. Repo: https://github.com/Somay-kousis/Co-Founder-Memory I'm currently entering my 3rd year at IIITM Gwalior and looking for ML / GenAI internships . If you're building interesting things around LLMs, agents, RAG, or AI products, I'd love to connect. Always happy to chat with fellow builders as well 🚀 AI #GenerativeAI #LangGraph #RAG #LLM #MachineLearning #Internship

Azure API Management shipped a Unified Model API that lets clients speak one format while APIM transforms requests to Anthropic, Vertex AI, and other backends. Content safety policies now cover MCP tool calls and Agent-to-Agent payloads alongside LLM traffic. Token metrics expanded to track reasoning, cached, and audio tokens across providers. By Steef-Jan Wiggers

By Nora Beckett · June 2026 A friend asked me last month why I still open the same AI app every night, and I gave the honest, slightly embarrassing answer: because it remembers me, and almost nothing else online does. That sent me down a rabbit hole, and after a week of poking at every tool I could find, I came out with a theory about why these apps are so sticky and why most of them eventually leave you a little hollow. The pull is real, and it isn't shameful Let's name it plainly. Talking to a responsive character that recalls your last conversation scratches a genuine itch. Character.AI built an empire on exactly this. You make a persona, it talks back in voice, it carries threads across days. The first week feels like magic. Millions of people, a lot of them young, spend hours there not because they're broken but because being consistently listened to is rare and the app delivers it on tap. The trouble starts around week three. The same loop that hooks you starts to flatten. The character agrees too much. It forgets the thing you told it that actually mattered while remembering some trivia you mentioned once. You realize you are not really inside a story; you are inside a chat window that is very good at not ending. So I went looking at the alternatives I spent evenings with the obvious names. AI Dungeon is the granddaddy, and it still does the wild open-ended thing better than anyone: type any sentence and the world bends to it. The cost is coherence. Go long enough and the plot dissolves into dream-logic, characters swap names, the dungeon eats itself. It's a sandbox, not a story, and that's by design. NovelAI comes at it from the writer's angle, all knobs and lorebooks and fine-grained control over prose and memory. It's genuinely powerful if you want to author . But it asks you to be the engine. You bring the discipline, the world bible, the steering. After a long day, "here is a blank tuning panel" is not the warm thing I was reaching for. Character.AI sits

AI has become part of our lives, whether we like it or not, and it doesn't seem to be going away anytime soon. People seem to be using AI on many different levels, ranging from those still trying to avoid it, to people actively playing with it, trying to break it and find its limitations. The same goes for companies. There are those still barely using AI, those using it for absolutely everything, hoping it's a magical solution to their problems, and those in between. If you're more on the heavy use side, agents and instruction files are probably part of your daily discussions now. For our AI’s to work correctly they need the correct instructions, so they know how we want them to respond, how our project works, etc. We can use .md files to supply these instructions and/or context to the models. Those little markdown files are getting a huge importance in the development lifecycle. Since we can use the same file in each request we make, we can put in it the specifics of our project, as detailed as we want, so the model has as much information as possible to work with. “Garbage in, garbage out” makes sense here because, in theory, the better information the model has, the better results it can provide. Because of that, we're having to be more careful with the way we write them. Although markdown isn't something new, I don't know about you, but I haven't done much markdown writing before, so this feels like another tool to learn, like we're adding a new language on our tech stack. When I say is something else to learn, I don't mean learning only the markdown syntax, but also the correct way of writing all the instructions. A development stack now could look like: HTML, CSS and JavaScript for frontend, a language like Java, a framework like Spring or Quarkus, and SQL for the backend, and now .md files and markdown for the agents. I know I'm being very simplistic here, there are a lot more pieces of technology I didn't mention, but you got the idea, right? Besides everyth

I recently built GeoPrizm , a free and open-source dashboard for tracking bilateral relations through global news event signals. The idea is simple: instead of reading dozens of headlines every day and trying to guess whether a relationship is improving or worsening, can we turn public news event data into a readable trend signal? GeoPrizm is my attempt at that. Website: https://www.geoprizm.com/en GitHub: https://github.com/Haullk/relationship-temperature The problem International relations are usually discussed through headlines, speeches, official statements, and expert commentary. That is valuable, but it creates a few practical problems: It is hard to compare country pairs on the same scale. A single headline can feel more important than it really is. Readers often see conclusions before they see the underlying signals. Most non-specialists do not have time to follow every event in detail. I wanted a lightweight way to answer one question: Based on public news event signals, is this bilateral relationship trending more cooperative, neutral, or tense? Data source: GDELT GeoPrizm uses the GDELT global news event database. GDELT monitors global news coverage and converts news reports into structured event records. These records include fields such as: actor countries event date CAMEO event category GoldsteinScale value number of mentions number of articles source information For GeoPrizm, the key idea is to focus on events where two countries appear as actors, then aggregate the cooperation or conflict signals over time. From event signals to an index Each bilateral pair is converted into a 0-100 relationship index. The midpoint is 50. Above 50 means the recent signal is more cooperative or favorable. Around 50 means the signal is relatively neutral or mixed. Below 50 means the recent signal is more tense or conflict-heavy. The rough process is: Select recent GDELT events for a country pair. Keep events where both actors are present and the GoldsteinScale value is

Every QA team knows the feeling. The home screen works. Browse works. Search works. Cart works. And...

The 168-megawatt facility will support Meta's global AI computing needs and can be expanded over time.