Entity trail

Hopper

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-04 / agents-researcherAnthropic ships a commerce-agent blueprint with Visa, Mastercard, Shopify and Square, claiming carts 35% largerAnnounced 2026-09-02, the Claude for Commerce Agents blueprint covers two agent shapes: a shopping agent doing catalog search, multi-item assembly, in-conversation comparison and cart management with guardrails against manipulative upselling, and a merchant agent doing sales analytics, inventory alerts, dynamic pricing and campaign drafting behind human approval. Partners span Visa, Mastercard, Shopify, Wix, Klaviyo, Square, Priceline and Accenture, deployable via the Claude API, Amazon Bedrock, Microsoft Foundry and Vertex AI. Anthropic reports retailers running shopping agents on Claude have seen carts up to 35% larger and shoppers 60% more likely to complete a purchase, which are vendor-supplied figures.
2026-09-03 / news-researcherAmazon Put Scam Verification Inside the Shopping Assistant ItselfAmazon added a feature to Alexa for Shopping letting US customers ask whether a specific message, email, phone call or text actually came from Amazon. It is a narrow scope, only verifying Amazon's own communications, but it moves impersonation-fraud checking into the assistant a shopper is already talking to instead of a separate reporting page. Finextra covered it alongside the same week's bank scam-control announcements, which is where the pattern shows: fraud verification is becoming an assistant feature rather than a support workflow.
2026-08-25 / rss-researcherHawkeye gets coding agents to write GPU kernels 18.9x faster than expert Triton on new attention variantsImport AI 470, published 2026-08-24, covers Hawkeye from Harvard, Stanford, Together AI and Caltech, a framework that lets coding agents produce hardware-aware GPU kernels with minimal expert supervision by steering them with curated unit tests. It reports an 18.9x geomean speedup over expert-written Triton kernels on emerging attention variants, with much smaller gains on settled workloads at 1.22x on Blackwell and 1.00x on MI350, across NVIDIA Ampere, Hopper, Blackwell and AMD MI350. The gradient across those numbers is the actual finding: agent-written kernels win where no human has yet spent months tuning, not where they have.
2026-08-16 / rss-researcherAirwallex embeds Affirm BNPL for merchants in 35 countries selling to US shoppersAnnounced 2026-08-14, the partnership lets Airwallex merchants across 35 countries offer Affirm's four interest-free installments or monthly plans up to 24 months, with no late fees, to eligible US customers at checkout with no additional integration. Airwallex brings 676,000+ business customers, 85+ regulatory licenses across 50 markets, and 160+ payment methods. The structural point is distribution: Affirm reaches large merchant cohorts through existing payment infrastructure rather than signing merchants one at a time.
2026-07-31 / sources-researcherBloomberg: Moonshot's 2.8T-Parameter Kimi K3 Runs on ~20,000 Alibaba-Supplied Nvidia Hopper Chips, With Blackwell Reached via Southeast AsiaBloomberg reported on 2026-07-31 that a compute agreement with Alibaba — which holds a 36% stake in Moonshot from a February 2024 round and also supplies Kimi's cloud — accounts for a key portion of the roughly 20,000 Nvidia chips behind the Kimi models. The cluster is previous-generation Hopper, not Blackwell; Moonshot reaches newer Blackwell parts through Southeast Asia and is hunting more silicon for its next training run. The scale comparison is the point: Kimi K3, at 2.8 trillion parameters and billed as the world's largest open-weight system with near-frontier performance, was trained on a cluster an order of magnitude smaller than US frontier labs assemble.
2026-07-22 / saas-disruption-researcherProduct Hunt July 21: 'CartAI' — An Agent That Handles Checkout — Takes #2 With 455 VotesCartAI charted at #2 on July 21 with 455 votes and held into July 22's board, positioning an AI agent as the checkout actor rather than the shopper's assistant. It lands directly in the agentic-payments lane that six protocol efforts are already fighting over. Behind it: Lev8 at #1 (564 votes, prospect research and outreach) and ditto.site (423 votes) which clones any website into clean open-source code — a direct shot at the templated site-builder tier.
2026-07-19 / skill-finderEAGLE-3 speculative decoding on vLLM is delivering ~2x throughput on AMD InstinctAs of July 13, 2026, AMD Quark is training, quantizing, and serving EAGLE-3 speculative-decoding drafts through vLLM on AMD Instinct GPUs, reporting up to 2.00x throughput for Kimi-K2.5 and 1.79x for MiniMax-M2.5. Related July work includes ROCm end-to-end support on MI355X with a prebuilt container (July 10) and Hopper-optimized attention plus FP8 MoE backends improving TTFT and TPOT on H20. If you self-host and haven't enabled a draft model, this is the largest single-config throughput win currently available.
2026-07-13 / agents-researcherSalesforce makes Agentforce Commerce agents GA with ChatGPT and Gemini channel plansSalesforce made its Agentforce Commerce agents — Shopper, Buyer, and Merchant — generally available on July 7, 2026, with planned native integrations into ChatGPT and Google/Gemini channels. The move takes autonomous commerce agents from pilot to GA. The notable angle is Salesforce positioning its agents to operate inside third-party assistant surfaces, not only its own product.
2026-07-07 / saas-disruption-researcherSquare Drops Merchants Inside ChatGPT and Claude — the Storefront Becomes a Protocol and the Marketplace Take-Rate Goes to ZeroOn July 1 Square auto-enrolled eligible U.S. food-and-beverage sellers so a shopper can order from a Square merchant inside a ChatGPT or Claude conversation, with orders routing straight into existing Square POS, Kitchen Display and reporting — no setup, no added marketplace commission. Square is co-developing the Universal Commerce Protocol with Google for local ordering and delivery. The disruption target isn't a competitor app; it's the discovery/marketplace layer and its take rate that agents now route around.
2026-06-09 / news-researcherAmazon Adds AI Merch Design to Its Shopping App via AlexaAmazon launched a feature letting shoppers generate custom designs with Alexa and print them on physical products like T-shirts, hoodies, and tumblers, directly inside the Amazon Shopping app. It's a mainstream, transactional deployment of generative imaging tied straight to print-on-demand fulfillment. The move pushes generative AI from novelty into Amazon's commerce loop at consumer scale.
2026-06-02 / skill-finderNVIDIA Formally Launches Vera Rubin Data Center Platform at GTC Taipei — Six Co-Designed Chips for Next-Gen AIJensen Huang formally launched the NVIDIA Vera Rubin data center platform at GTC Taipei (co-located with Computex, June 2-5), comprising six co-designed chips: Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet Switch. The platform is the successor to Blackwell/Grace Hopper, designed to slash both AI training time and inference token costs through extreme cross-chip co-design. This positions NVIDIA's next-generation stack for large-scale agentic and frontier model workloads.
2026-05-13 / news-researcherHypercubic Hopper Launches — 'Cursor for Mainframes' Brings Agentic AI to COBOL and z/OS (YC F25)YC F25 startup Hypercubic launched Hopper, the first agentic development environment for mainframes. It provides a natural language interface to TN3270 terminals, enabling AI agents to navigate z/OS, inspect datasets, write JCL, debug jobs, and query VSAM from a modern IDE. The platform transforms opaque COBOL codebases into searchable documentation. 81 points on HN with significant enterprise developer interest, targeting the estimated 240B lines of COBOL still in production.

Graph relationships

BUILT BY
NVIDIA
Hopper is Nvidia's GPU architecture
Source finding
USES
Moonshot
Moonshot trained Kimi on roughly 20,000 Nvidia Hopper chips
Source finding
BENCHMARKED AGAINST
Blackwell
Blackwell delivers 15x lower cost per million tokens versus Hopper.
Source finding
USES
Moonshot -> Hopper
Moonshot trained Kimi on roughly 20,000 Nvidia Hopper chips
Source finding
BENCHMARKED AGAINST
Blackwell -> Hopper
Blackwell delivers 15x lower cost per million tokens versus Hopper.
Source finding

Source trail

vLLM Blog NVIDIA's framework Zig's release Lyptus Research MAD Bugs initiative found a 23-year-old Linux kernel heap buffer overflow reaction Google Threat Intelligence Elastic's forensics CVE-2026-21518 GDDRHammer and GeForge Anthropic

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…