The Archive

Distributed deep learning depends on fast, reliable GPU-to-GPU communication using the NVIDIA Collective Communication Library (NCCL). When training slows down,... Distributed deep learning depends on fast, reliable GPU-to-GPU communication using the NVIDIA Collective Communication Library (NCCL). When training slows down, it becomes challenging to determine why and what to do next. A problem can span computation, communication, a specific rank, or underlying hardware. NVIDIA NCCL Inspector accelerates triaging by providing a lightweight and continuous… Source

Reddit discussion about ECCV reviewer citing author's own prior arxiv work without attribution, requesting comparison to their unpublished version.

DeepMind's AlphaEvolve uses Gemini to build a coding agent demonstrating scalable impact across multiple domains.

Developer reports Reddit ban for using Codex in a Product Hunt launch, citing AI-generated content moderation policies.

Startup Battlefield 200 applications are open, but only for three more weeks. Apply by May 27 for your shot at VC access, global visibility, TechCrunch coverage, $100K equity-free, and more opportunities for major scaling impact. Pre-Series A founders — and anyone who knows a startup worth backing — this is your reminder: the deadline is approaching fast, and the […]

So ya... quite straigh forward again... if Anthropic would have people to help one when there's an issue instead of a mindless bot... less people would have to come here to complain. Since as it stands this is the only possible outlet.

If startup visibility, traction, and real deals matter, you should be on the TechCrunch Disrupt 2026 exhibit floor. Get your 6' exhibit table before your competitor does.

Self-driving has been “almost here” for over a decade. But somewhere between DARPA challenges and a handful of driverless trucks hauling freight between Dallas and Houston, Aurora co-founder and CEO Chris Urmson’s story changed. The self-driving truck company started commercial driverless operations last April and is now scaling from a handful of trucks to hundreds this year. On this episode of TechCrunch’s Equity podcast, we’re bringing you a […]

Satirical Reddit post imagining EU-regulated Claude with regulatory taxes and compliance overhead.

Save to Spotify is a new command-line tool designed specifically for AI agents like OpenClaw, Claude Code, or OpenAI Codex. If you're the kind of person who collects research on a topic, then feeds it through their AI of choice to create audio summaries and personal podcasts, this lets you save them right alongside the latest episode of The Vergecast and Welcome to Night Vale on Spotify. To set it up, you need to download and install the Save to Spotify CLI from GitHub. Then you just prompt your AI agent as normal, but tack on "and save to Spotify," and it should show up right in your podcast...

The last 2-3 days, Claude has gotten really dumb. I have to switch over to ChatGPT to get anything done. I threw it the same file it spat at me and explained the problems and it continued to waste my usage going back and forth with itself. "It could be this... no, that's not it" type stuff. 3,022 tokens gone before I interrupted it. Threw it into ChatGPT and it was like "Here's your problem" and fixed it. It seems like Claude does this every once in a while... just gets dumb for some reason. Like, it will go for a super long time like Einstein and then a couple of weeks later gain an e...

The $100 Fitbit Air is available for preorder today.

The Fitbit Air can be preordered today and will be available starting May 26th. | Image: Google It's a Whoop dupe. That was my first thought when I saw the new $99 Google Fitbit Air. You can hardly blame me. The band is screenless with a metallic fabric clasp. My eyes flickered between the Fitbit Air and my wrist, where I'm wearing a Whoop MG. Was I not seeing double? But as my press briefing went on, my opinion started changing. The Air is sort of like the OG Fitbits that Whoop then duped once Fitbit went all in on smartwatches. Think back to 2012, when the Fitbit One could clip to your pant...

Reddit user seeks help understanding an unspecified OpenAI update; no technical details or claims provided.

AMD announces Instinct MI350P PCIe accelerator with CDNA 4 architecture for AI inference and training workloads.

Moonshot's annualized recurring revenue topped $200 million in April, driven by rapid growth in paid subscriptions and API usage.

User complaint about Claude's conversational style and unsuccessful attempts to modify it via prompting.

A few weeks ago, I shared some tips on my Claude Code workflow. In the comments, quite a few people mentioned that they were burning through their tokens super fast and tbh I could totally relate. This is something I particularly struggled with at the beginning, which pushed me to take a closer look at it. Turns out most of my token usage wasn't coming from Claude's answers, but from the setup. Things I actually use: * **Start a new chat for unrelated tasks.** Every message in a long conversation resends the full history. That's not obvious until I realize a 40-message thread is burning tok...

Musk was “prepared to do the for-profit, provided he would get control.”

Reddit speculation about Anthropic-SpaceX compute partnership; unverified claim presented as meme commentary.

OpenAI launches GPT-5.5 and GPT-5.5-Cyber under Trusted Access program for verified security researchers to accelerate vulnerability discovery.

Users will be able to create a podcast from Codex or Claude Code and import it to Spotify

Unverified Reddit post claiming access to leaked private messages about Sam Altman's 2023 firing.

Spotify's AI DJ feature now supports French, German, Italian, and Brazilian Portuguese.

Well I might even get 2 days work done this week with Claude the fraud. 12 hours in and 52% of my usage gone (off peak no mcp). That sux !

The compute is real. The implications are stranger than the headline suggests. Colossus 1 which is 220,000 Nvidia GPUs, 300+ megawatts, is now running Claude inference. Anthropic moved fast: Claude Code limits doubled overnight, peak-hour caps removed, Opus API rates up. For anyone who's been hitting walls, this is immediately tangible. But the deal deserves more scrutiny than it's getting. Musk included a clause reserving SpaceX's right to reclaim the compute if Claude "engages in actions that harm humanity." That's not standard infrastructure boilerplate. That's a kill switch written int...

Reddit discussion on OpenAI's aggregate energy consumption metrics; data-focused, no ethical framing.

Xiaomi releases Mimo v2.5, a 310B sparse MoE multimodal model with 1M token context supporting text, image, video, and audio.

Parloa uses OpenAI models to build voice-driven customer service agents with simulation and real-time deployment capabilities for enterprises.

DeepSeek approaching $45B valuation in Series A led by China's Big Fund, signaling major state backing for domestic LLM competitor.