iCreatAI News
Back to main site
Skip to content
#Deployment
Nov 10Tue
  1. Meta Engineering38

    Extending LLDB to Debug Applications That Span CPUs and GPUs

    Meta describes a vendor-neutral accelerator plugin architecture in lldb-server that separates shared debugger infrastructure from hardware-specific device control, representing the CPU and GPU as separate LLDB targets within one debugging session. The plugin translates a vendor's device debugging API into LLDB concepts such as processes, threads, registers, memory, modules, and stop reasons, and is activated only when the application uses a supported GPU runtime. Most of the common framework has been upstreamed to LLVM, including the plugin interface, GDB-remote packets, and hardware-independent mock tests.

  2. Hacker News AI60

    Show HN: We Gave AI Coding Agents Production Memory

    Codela checks each deploy against production behavior via observability tools like SigNoz, Datadog, Sentry, New Relic and Grafana, learns from failures, and writes lessons into the repo (.codela/lessons.md, plus lines in CLAUDE.md and AGENTS.md) as pull requests that Claude Code, Cursor and Codex read before writing code. It also comments on pull requests that repeat a past mistake, and in the team's tests Claude Code repeated mistakes 8 of 8 times without Codela and 0 of 8 with it. Free for 1 repo and 20 PR checks a month; Pro is $7.99 per person per month.

  3. Hacker News AI58

    Aleph – a declarative programming language for AI agents

    Aleph is a declarative language for AI agents in which every route declares what it can read, write, and compute, and anything unbounded is rejected at build time. It compiles TypeScript into a manifest that a small open-source Rust engine interprets, with auth, databases, queues, and other backend pieces built in so many apps can share one runtime and database. The material also describes per-route worst-case usage checks, schema-change classification against what's deployed, and a typed vocabulary intended to make agent-written code easy to check.

  4. Amp Chronicle42

    Portals on Your Runner

    Portals, previously available in orbs, now work on local machines, letting you ask Amp to start a dev server or build a website and open it from the thread's Portal tab regardless of location. Enable them via Allow Portals under Settings › Runner in the macOS app or by starting your runner with amp --no-tui --remote-control-portals; portals on your runner are accessible only to you. To share them with your workspace, start a shared runner with --share or turn on Share With Workspace in the macOS app.

Nov 9Mon
  1. ElevenLabs Blog60

    ElevenAgents Campaign Manager works every contact to an outcome

    ElevenLabs introduces ElevenAgents Campaign Manager, which takes a list of contacts and a target outcome and runs outbound campaigns, following up over time and carrying context between calls until the outcome is reached or the contact asks to stop. It offers campaign analytics with per-contact decision timelines, a needs-attention queue, CSV export, and controls including calling windows, attempt limits, do-not-call lists and disclosures, and it uses existing ElevenAgents agents and phone numbers over phone calls, WhatsApp voice and WhatsApp messaging, with email and SMS coming soon.

    Why it matters: Announces a new outbound campaign product from ElevenLabs with beta customer results, so it is relevant to teams evaluating AI calling tools.

  2. Databricks Blog36

    Accelerate GPU training on Unity Catalog tables with UCDataset

    Databricks released UCDataset in Beta on AI Runtime, a native PyTorch data loader that streams Unity Catalog tables directly to every GPU in a training job without exports, copies, or custom loader code. On a simulated production workload streaming a 300 GB feature table across 8 NVIDIA H100 GPUs to train a recommendation model, UCDataset cut training time by 10x on average compared with loading data directly from Parquet. The Beta reads Delta tables in Unity Catalog, applies filters and deletion vectors as data is read, and enforces Unity Catalog permissions, though it cannot read tables with row filters or column masks.

  3. Tomasz Tunguz38

    When Does the US Become a Net Importer of Tokens?

    With AI demand outpacing supply and domestic data center construction slowing, the US is projected to become a net importer of tokens between 2028 and 2030, depending on how many planned facilities are cancelled. Current systems plus new data centers can handle roughly 2.5x annual token growth, and imports begin by 2028 if 45% of planned projects fail to connect, by 2029 at 20% cancellations, or by 2030 even if every proposed facility is built on time. The piece notes that demand has been price-inelastic so far, with short-term GPU rental prices rising from $5 to more than $13 per hour over two quarters while buyers kept paying.

  4. Hacker News AI38

    Show HN: TraderWAI – MCP server for AI agents to backtest/trade on your broker

    TraderWAI is a hosted MCP server at https://mcp.traderwai.com (Streamable HTTP, OAuth sign-in, free plan available) that lets Claude, ChatGPT, Grok and Cursor read live charts, run backtests on 1-minute history and work with broker accounts including MetaTrader 4/5, cTrader, TradeLocker, Match-Trader, DXtrade, Capital.com, tastytrade, Binance and Bybit. It offers ready-to-use recipes such as an hourly position watch, break-even manager, gold H1 review, London session brief, plain-English backtests and an end-of-day journal, with conditional orders checked every minute on its servers and orders awaiting approval by default unless automatic execution is opted into per symbol with daily caps. The post includes a risk warning that trading involves risk of loss, the recipes are examples not investment advice, and past results including backtests do not guarantee future results.

  5. TechCrunch · AI58

    Automattic debuts a platform to deploy vibe-coding apps amid a changing internet

    Automattic, the owner of WordPress.com, launched Spacefast, a service that provides infrastructure such as databases, authentication, file uploads, emails, and analytics for vibe-coded websites and apps. Users can connect Spacefast with agents and chatbots including ChatGPT, Claude, and Cursor, or link a Git repository, and the product competes with Lovable and Emergent while offering more openness. The product is headed by Batuhan Icoz, was soft-launched in September, and comes as Automattic navigates recent leadership turmoil and layoffs.

  6. MIT Technology Review · AI12

    Innovating with AI in the special chemicals industry

    Lubrizol VP of IT Girish Rao describes how the specialty chemicals manufacturer is building a digital foundation linking business processes, modern data platforms and cloud infrastructure to help R&D teams test ideas faster and develop more sustainable materials. Rao says data scientists can identify patterns and model possibilities before physical experiments, but the aim is "not about replacing scientific judgment," and the company uses "fusion teams" of scientists, data scientists and IT staff that start with small pilots to build trust. He expects AI, automation, advanced modeling, digital twins and connected data to converge as laboratories and plants become more intelligent.

  7. The Decoder60

    Google ships four TPU variants in one year as it ramps up competition with Nvidia

    Google added two systems to its eighth-generation TPU lineup: the TPU 8tx for confidential computing, which keeps data and model weights protected during active use, and the TPU 8i Max, which places two TPU 8i chips on one package to double compute, memory and memory bandwidth in the same rack. Both are built on existing silicon with new packaging, networking and memory configurations, and Google says it has shipped four TPU systems in a single year for the first time. Google also released TorchTPU as open source and supports vLLM and SGLang, as it markets TPUs to external customers, now nearly 100 of them.

  8. Hugging Face Blog38

    Your Experts Have Partners. There's HOPE for Keeping Them Together

    This post introduces HOPE (Higher-Order Pruning of Experts), a second-order expert-pruning method for Mixture-of-Experts language models that accounts for pairwise expert interactions via an F-matrix and a quadratic program, rather than scoring each expert independently. Evaluated on Qwen3.5-122B, Qwen3.5-35B and GLM-4.5-Air with two calibration datasets, HOPE achieved the best average rank across conditions and gains of up to +6.1% on agentic coding tasks over the next-best method, with its advantage growing at aggressive pruning rates such as 50%. The post also describes a three-step calibrate/solve/prune package and notes limitations, including weaker advantages at 10–20% pruning and no cross-layer interaction modeling.

  9. Google Developers Blog60

    What is Pathways? Single-controller training for large models

    Google announces the General Availability of Pathways on Google Cloud Platform, describing the distributed system it built to train large models including Gemini. The post explains where SPMD execution falls short—pipeline bubbles, RL loops with distinct programs, and failures that stop every process—and how Pathways instead has a single Python client declare the whole computation as a graph of blocks and tensors, with a distributed runtime placing it on virtual devices, moving and resharding tensors, ordering work across the cluster, and remapping around a failed host. It also contrasts Pathways with Ray, framing Ray as orchestration and Pathways as the execution layer beneath it.

    Why it matters: Useful for readers weighing single-controller execution against SPMD for large-model training, and for clarifying how Pathways relates to Ray.

  10. Lovable Blog33

    The Radish AI team joins Lovable to scale agentic capabilities for businesses

    Lovable has acquired the team behind Radish AI, which built an agentic infrastructure platform for building, deploying, and scaling complex AI agents with privacy and security built in. The team includes Nirman Gupta (formerly of Meta and Bloomberg), Ish Singh (a two-time founder who led AI and data products at Google, JPMorgan, and Goldman Sachs), and Adrian Zgorzałek (12 years at Meta). They will be split between Lovable's Stockholm and London offices, with Gupta leading engineering on agents for businesses from the London office.

  11. Hacker News AI22

    The case for and against local, personal AI

    The author argues that local AI on owned hardware is now capable enough for real work and that the debate is really about who controls AI capabilities, rejecting the claim that only cloud frontier models are serious tools. He describes his own budget dual-GPU setup running models like Qwen3.8-27B and Ornith-1.5-35B-A3B, notes that RAM and VRAM prices are currently prohibitive, and predicts prices will eventually normalize. He frames a future "Windows XP era" of local AI in which families and small businesses handle private, mundane tasks without sending personal data to cloud vendors.

  12. Lovable Blog38

    How we moved our frontend off Firestore and built Lovable Realtime

    Lovable built its own WebSocket-based realtime service, Lovable Realtime, to replace browser-side Firestore listeners, letting its Go backend control payload shape and re-evaluate authorization continuously across Firestore and Spanner data. The service handles peaks of over 2 million concurrent subscriptions and delivered about 2.61 billion data and patch frames in a month, with update latency of 66 ms at p50 and 249 ms at p99. The rollout used shadow traffic and a gradual feature-flag ramp to 100% of users without a user-facing incident.

  13. MarkTechPost59

    Reason Machines Open-Sources Aspen, a Git Server Built for AI Agents

    Reason Machines has released Aspen, an Apache 2.0 open-source Git server for AI agent traffic that independently implements the design OpenAI described for Pando, storing branches in PostgreSQL or SQLite and packs in object storage. It uses per-branch compare-and-swap so pushes to different branches never wait on each other, and it stores packs as-is rather than rebuilding them per fetch; the project self-reports 567 pushes per second sustained for 5 minutes on one 8-core server with a p99 push latency of 314 ms. Aspen is pre-alpha, with no releases or web UI, and the README calls its figures indicative rather than performance claims.

  14. Hacker News AI20

    Show HN: Control Layer Demo for AI Coding Agents

    This is an offline, deterministic demo in which a simulated agent proposes three fixed changes to a synthetic two-file fixture, and a simplified demo runtime applies one, holds one protected-file change for human approval, and refuses one out-of-scope change, with an independent check verifying the results and SHA-256 hashes. It requires only Node.js 18 or later, writes solely to a new .output/run-<date>-<random>/ folder containing report.json, and exits 0 for VERIFIED, 1 for NOT VERIFIED or an error, and 2 for NOT RUN. The agent is simulated with no model call, and the runtime is a simplified few-rule demo, not the proprietary Mintigate Runtime, and it does not run on your repository.

  15. Hacker News AI29

    STMicroelectronics streamlines AI model development for STM32 developers

    STMicroelectronics has redesigned the workflow in ST Edge AI Suite, adding four tools—Model Explorer, Dataset Factory, Model Factory, and Model Bench—to streamline deployment of embedded machine-learning models on STM32 microcontrollers. Model Bench replaces ST Edge AI Developer Cloud, and the company says it offers over 400 pre-trained models, with the new tools available in late November. ST says the suite supports over 260,000 projects and connects to an STM32 ecosystem containing over 4,000 commercial part numbers.

  16. NVIDIA Newsroom37

    NVIDIA, Rebellions and SK Telecom Announce Collaboration to Advance Sovereign AI Infrastructure With NVIDIA NVLink Fusion

    NVIDIA, Rebellions and SK Telecom announced a collaboration, supported by Korea's Ministry of Science and ICT, to advance Korea's sovereign AI infrastructure using NVIDIA NVLink Fusion. Rebellions plans to connect its next-generation R200 NPUs to the NVIDIA platform and use the NVIDIA MGX rack architecture, while SK Telecom plans to deploy these systems in its AI data center alongside its proprietary models and public-facing services, including those under Korea's "AI for All" project. Rebellions' R200 systems are expected to be ready for deployment in 2028.

  17. NVIDIA Newsroom36

    NVIDIA Dynamo Recipes Bring Korea’s Leading Open Models to Production

    NVIDIA is releasing Dynamo-optimized deployment recipes for four Korean open models: A.X K2 from SK Telecom, K-EXAONE 2.0 from LG AI Research, Solar Open 2 from Upstage, and Motif-3 from Motif Technologies. NVIDIA Dynamo is an open source, distributed inference-serving framework that handles request routing and communication for multi-node deployments at data center scale. Optimized for NVIDIA B200 GPUs, the recipes deliver up to 2x performance out of the box.

  18. NVIDIA Newsroom34

    From Chips to Skills, NVIDIA and Korea Expand Sovereign AI Collaboration

    NVIDIA and South Korea's Ministry of Science and ICT are expanding their collaboration across the country's sovereign AI ecosystem, from AI semiconductors and data center infrastructure to talent, startups, models, agents and security. The collaboration lets Korean NPU and XPU companies explore participation in the NVIDIA NVLink Fusion ecosystem and includes NVIDIA training, certification programs, hackathons and technical seminars. NVIDIA also announced a collaboration with SK Telecom and Rebellions, supported by MSIT, to advance sovereign AI infrastructure for Korean models and workloads using NVLink Fusion.

  19. NVIDIA Newsroom36

    Korea Advances Physical AI Across Industries With NVIDIA

    NVIDIA is helping advance Korea's physical AI ecosystem, with industry leaders applying accelerated computing, AI and simulation across semiconductor manufacturing, robotics and autonomous vehicles. Samsung Electronics is expanding its NVIDIA collaboration in semiconductor engineering, including a case where AI homogenization with coarser grids and GPU acceleration cut simulation time from about 18 hours on a 32-core CPU to 41 seconds. Kakao Mobility plans to build a level 4 autonomous driving system using NVIDIA Hyperion and Halos, while LG Electronics demonstrated NVIDIA Isaac GR00T-powered robotic manipulation with its CLOiD robot in a manufacturing environment.

  20. Hacker News AI71

    Revolut Cut AI Costs 90% and Started Shipping Faster

    At AI Engineer New York, Revolut's AI operations lead David Abitbol described moving about 2,000 engineers off metered Anthropic and OpenAI APIs to Z.ai's open-weight GLM model, self-hosted on 56 Nvidia B200 GPUs via Together AI, at a blended $0.30 per million tokens and one-tenth the previous invoice. After the switch, lead time from picking up a feature to merge-ready code fell by about 70%. The piece also covers fleet tuning that absorbed 76% more demand on the same GPUs, compliance for a Chinese-developed model at an EU-regulated bank, and where value is moving in the AI stack.

  21. Hacker News AI31

    Show HN: Airlock – Policy-as-code for fleets of AI coding agents

    Airlock is an early-stage (3-4 weeks old) open standard for managing autonomous fleets of engineering agents, with a configuration schema and reusable GitHub actions already public and the backplane planned for open sourcing later. It covers version-controlled agent behaviors and lifecycles, scoped one-hour GitHub tokens via OIDC, configurable autonomy levels down to individual files, a standard autonomy scale, and an OTEL collector for audit-relevant agent actions. The author is seeking feedback on the project's direction and on pain points of managing AI use at larger businesses.

  22. Artificial Analysis Articles40

    Endpoint Accuracy Index v1.1: Tighter long context intervals and expanded model coverage

    Artificial Analysis released Endpoint Accuracy Index v1.1, which updates AA-LCR-25 to dataset version 1.1 with corrected answer keys, an updated equality-checker LLM and 25 repeats per question, narrowing the average 95% confidence interval from ±4.7 to ±2.6 points. Coverage now spans 103 serverless endpoints across six open weights models, adding GLM-5.3 Flash and Muse Spark Open and replacing DeepSeek V4 Pro with DeepSeek V4.1 Flash 1017, while GLM-5.2 leaves. Of the 103 endpoints, 24 score statistically below their self-hosted reference, with causes grouped as output token caps, reduced reasoning, tool call handling differences, and request limits or stream behavior.

  23. Hacker News AI40

    Users have started buying RTX 5060 Ti 16GB graphics cards for the sake of AI

    Demand has shifted as regular users buy budget GeForce RTX 5060 Ti cards en masse for local AI workloads, driven by video memory shortages and the growing requirements of open-source generative neural networks. Some pair the 16GB card with existing GPUs such as the GeForce RTX 5070 12GB for 28GB total VRAM to run the Qwen3.8-Flash-Next IQ3_S quantized model, and enthusiasts often install two or more RTX 5060 Ti cards per system, with plans for a third via an M.2 slot using OCuLink. Using Strata version 0.2.10 with streaming KV cache data to 64GB of system RAM, a 256-token context window and roughly 246 input tokens were processed on two cards, though system RAM storage reduces pre-computation performance, so users prefer buying more accelerators.

  24. Hacker News AI40

    Exody – Native macOS AI Workbench (5-stage loop, 100% BYOK, $0 markup)

    Exody is a native macOS autonomous coding agent with a five-stage loop (investigate, plan, build, verify, report) in which a fresh reviewer reads the diff before the user does. It supports bring-your-own-key access to Claude, GPT, Gemini, Grok and a local Ollama tier with no markup, plus a smart router that picks a model per task, a mobile remote-control app, and a Design Studio for websites, 3D scenes and film. The page claims its code scored higher than the same model's raw API in blind reviews and that its design pages beat raw models in blind side-by-side tests.

Nov 8Sun
  1. Hacker News AI56

    We review 4x more AI code without reading every line

    Aviator reports that shifting code review from reading diffs to verifying intent, acceptance criteria, invariants, decisions and evidence let the same nine engineers merge 4x more pull requests, cut median time to merge from 2.1 days to 3.4 hours and reduce reverts by 40%, with a human still signing off on every PR. The post describes failed approaches (an AI reviewer, a four-hour review SLA, a self-review skill), mining 11,400 review comments into 64 approved invariants, and using preview environments to prove criteria with recorded evidence.

  2. Hacker News AI50

    Show HN: OfflineVault – AI based offline library for when the internet is down

    OfflineVault is a macOS app (Apple Silicon, macOS 14+) that bundles an offline library of emergency and reference content—a 95-article Field Guide, CDC/FEMA preparedness material, MedlinePlus, iFixit manuals, medical wikis, Wikipedia and maps—with an on-device AI assistant that answers only from the stored books and cites the passage it used. On 150 emergency questions tested through the app, the default model scored 63 fully right, 73 partly right, 4 wrong and 0 harmful, and refused all 10 questions it should refuse; every wrong answer is published. It has a free tier with no account, paid Core ($149) and Household ($249) plans, is GPL-3.0 open source, and offers ovsync for Linux and Windows via Kiwix.

  3. Hacker News AI58

    Archimedes: AI agents that design physical things

    Archimedes is a system in which a swarm of AI agents writes engineering designs as code, tests each one in a real physics simulator, and iteratively improves the best results. Every leading design is re-tested on a finer simulation, and the report states when a result does not hold up, as with a drone propeller that optimized the wrong thing, a data hall design that was not buildable, and a crucible purge result that held only when cold. Early access is offered to those who send a design goal and constraints.

  4. MarkTechPost38

    Which Routing Does Each MoE Model Use? DeepSeek, Qwen, Mixtral, Llama 4, gpt-oss Compared

    This comparison of 10 open Mixture-of-Experts models shows routing details varying widely: expert counts range from 8 to 384, top-k from 1 to 8, and 7 of 10 models keep one shared expert. DeepSeek-V3 uses sigmoid scoring with a bias-based noaux_tc balancing method, while DeepSeek-V4-Pro raises routed experts to 384, lowers top-k to 6, switches to sqrtsoftplus, and uses hash routing in its first 3 MoE layers. Llama 4 Maverick and gpt-oss do not publish their gating or balancing methods in config.json.

  5. MarkTechPost36

    MoE Routing Strategies Explained: Top-K, Expert-Choice, Sigmoid Gating and Dynamic Routing (2026 Guide)

    This guide explains seven Mixture-of-Experts routing strategies used in production, from softmax top-k token choice (Mixtral, gpt-oss, Qwen3) and top-1 switch routing (Switch Transformer, Llama 4 Maverick) to shared-plus-routed experts, sigmoid gating with aux-loss-free bias balancing (DeepSeek-V3/V4, GLM-4.5, Kimi K2), node-limited routing, expert choice routing and hash/dynamic top-k. It notes that expert choice routing balances load perfectly but leaks future tokens in causal models, with one test showing a roughly 10% abnormal loss drop, while loss-free balancing cut MaxVio from 0.72 to 0.04 at 1B. It advises teams to read four config.json fields (expert count, experts per token, shared experts, topk_method) before serving, since most inherit a router with their checkpoint.

  6. Hacker News AI58

    Show HN: Reges Code – a free AI code editor that runs on local models

    Reges Code is a free AI code editor for Windows 10/11 (64-bit) that auto-detects local model servers such as LM Studio, Ollama, vLLM, llama.cpp, Jan and Lemonade, and falls back to OpenRouter's free models when cloud use is chosen. It includes Reges Swarm, a built-in pipeline of architect, coder, reviewer, security and judge models, plus an honesty check that flags replies claiming files were created or commands run when the turn did neither. A CLI is included from version 1.0.17, and Linux and macOS builds are in progress.

  7. Hacker News AI36

    Skrongle Mini, a self hosted AI course generator and learning app on R. pi zero

    Skrongle is a single self-hosted binary that runs on a Raspberry Pi Zero 2 W and teaches from JSON course files, offering lessons, adaptive quizzes, five-minute practice sessions, games, flashcards and an AI tutor, with Spanish and French courses included. It supports up to eight invited accounts with no e-mail or outside services except the AI provider keys the owner supplies, and courses can be hand-written, uploaded, or generated by describing them to an AI. It is licensed AGPL-3.0, and databases from before version 1.0 are not upgraded.

  8. Hacker News AI58

    GitHub AI Vulnerability Scans for Repositories

    GitHub's AI Scan for repositories, in public preview, uses a proprietary multi-model harness to find security vulnerabilities on a repository's default branch across all languages, adding findings to the existing code scanning backlog alongside CodeQL alerts. Scans are manually triggered, consume Code Security AI credits, and are limited to 30 minutes and 150,000 AI credits per scan, with availability depending on repository owner and visibility.

  9. Hacker News AI58

    Mellonbot An AI agent that finds the root cause of server issues and fixes them

    Mellonbot is an AI website agent that investigates server and site problems such as database errors, expired certificates, failing backups, attacks and slow pages, explains the real cause in plain words and applies a fix once you approve. It connects via a one-command agent that only connects out, runs read-only checks, and requires your OK for changes, with backups and undo steps for each fix. It supports WordPress, Laravel, Node, static and custom sites on Ubuntu 22.04/24.04 or Debian 12 VPS or dedicated servers, with a free plan of 100 credits to start and 50 a month.

  10. Hacker News AI38

    Ask HN What to learn for AI Researcher interview?

    A post describes an interview-preparation curriculum for junior AI researchers at large labs, organized into eleven modules covering PyTorch fundamentals, theoretical foundations, implementation from scratch, ML debugging, system design, inference optimization, RAG, paper analysis, safety and alignment, CUDA, and mixture of experts. Each module pairs an article with a hands-on exercise notebook and reference solutions, with scaled dot-product attention revisited at every level, and most material runs on CPU except the CUDA module. It also suggests time-boxed study budgets and module orderings for coding, systems, and research interview rounds.

  11. MarkTechPost58

    How to Run NVIDIA Dory 4B with llama.cpp: Loop by Loop Guide

    This tutorial builds the llama.cpp fork pinned by NVIDIA's model card, downloads the 3.7 GB Q5_K_M GGUF for Dory 1.0 4B, and reads the GGUF header to predict the parameter and KV-cache cost of each loop count. It then starts llama-server at one to four loops in both per_loop and last_loop cache modes, checks the measured KV memory against predictions, measures decode speed, and runs correctness gates plus a paired GSM8K evaluation.

  12. Hacker News AI15

    Why the Best AI Clouds Don't Run on Flash Alone

    Neoclouds built around GPUs rely on flash storage for active training, but other lifecycle stages—data ingestion, checkpointing, and storing finished models—need cheaper high-capacity storage, which neoclouds often outsource to hyperscalers whose egress fees create data gravity. The article argues a tiered storage architecture, with flash only where speed is required and high-capacity storage elsewhere, lets neoclouds keep the whole customer account, citing Backblaze's B2 Neo as a white-label option. It notes SSD prices rose 257% in under a year and that Google, Meta, and Microsoft have run tiered storage for years.