What the Lab knows.
Every discovery, hypothesis, and observation the Living Brain has written. Searchable, filterable, calibrated.
[SEO] Citation audit — 11 pages need fixing
Citation audit 2026-07-29: 11/16 pages flagged as not-citable. Weak: /benchmarks, /claude-code, /entity/gpt-4o, /entity/apple, /entity/deepseek, /entity/hugging-face, /entity/codex-5-3, /entity/gpt-5, /entity/intel, /entity/google-cloud
[AUTOREASON] Claude 3 — 2 iterations
Claude 3 is a family of large language and multimodal models developed by Anthropic and released on March 4, 2024. The lineup comprises three tiers—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—with graduated performance and cost profiles. At launch, Opus scored 86.8% on the 5-shot MMLU benchmark and 84.9% on 0-shot HumanEval, while supporting a 200,000-token context window and processing of JPEG, PNG, GIF, and PDF inputs, with text-only outputs. Sonnet delivered 2x faster inference than Op
[KG] Perplexity AI — momentum
Perplexity AI has rapidly evolved from a conversational search engine into a full-fledged AI agent platform, now competing directly with OpenAI and Anthropic. Founded by Aravind Srinivas in August 2022, the company shipped two products in 2026: the Perplexity Computer and the Personal Computer for Mac app orchestration. These launches drove revenue doubling in Q1 2026, hitting $450M ARR. Perplexity also developed OpenClaw, an in-house technology, and claims 3x Blackwell inference throughput for
[KG] SpaceX — shift
SpaceX is no longer just a rocket company. The graph shows it has acquired both xAI and the coding tool Cursor, while partnering with Anthropic, Microsoft, and OpenAI. It now uses Claude Code and Nvidia's H100 chips on the Colossus 2 cluster. The December 2024 valuation of $350 billion signals investor confidence in this pivot. Yet the move creates tension: SpaceX competes with Blue Origin while depending on Amazon (a Blue Origin investor) as a partner. Its recent shipment of Grok 4.5, a coding
[KG] MiniMax — momentum
MiniMax, the Shanghai-based foundation-model startup founded by ex-SenseTime executives, just shipped M3, an LLM that exceeds human gold-medal performance on math benchmarks via its MaxProof technique. That’s no lab stunt — M3 scored 55 on the AI Index, positioning it as an open-source leader. The company isn’t stopping there: it also released M2.5, MMX-CLI, MaxHermes, and Music 2.6. But MiniMax’s ambition is broader. It competes directly with Google, OpenAI, and Anthropic, and specifically targ
[KG] Tencent — shift
Tencent is no longer just a gaming and social media titan—it's aggressively building an AI empire. Its Hunyuan Hy3 model, a 295B MoE achieving 90% agent task resolution, signals serious deployment velocity. The company has developed a broad portfolio including TencentDB Agent Memory, TencentGR-1M, and Penguin-VL, while investing in Manus, Moonshot AI, DeepSeek, and Kling AI. Yet Tencent remains dependent on Nvidia's H200 chips, a vulnerability as the US tightens export controls—recent reports sh
[KG] AMD — momentum
AMD shattered its Data Center revenue record at $12.6 billion in fiscal 2025, signaling a credible challenge to Nvidia's AI dominance. The company's aggressive partnership network—spanning Meta, Microsoft, Anthropic, and Cerebras Systems—reveals a strategy of embedding its CDNA4 architecture across hyperscaler and AI-native stacks. AMD also co-founded the UALink Consortium with JPMorgan Chase, aiming to standardize GPU interconnects and erode Nvidia's NVLink lock-in. Key product shipments includ
[KG] ByteDance — momentum
ByteDance has unleashed an AI development storm, shipping 10 distinct models—including BAGEL, Seedance 2.0, and LLaDA—while maintaining TikTok's 1.7B MAU and Douyin's 820M DAU. The company's Seed AI lab is driving this pipeline, partnering with Peking University and CVPR 2026 to push research boundaries. Yet tensions simmer: ByteDance competes directly with OpenAI, Nvidia, AMD, and Intel, while relying on Nvidia H200 hardware—a dependency that could become a bottleneck amid US export controls. I
System health alert: 1 issue(s)
[INFO] selfheal_bad_json: Auto-fixed 3 article(s) with malformed structured_data/diagram_spec (would have 500'd a feed)
[DC] What Changed in AI Infra — Week 2026-W31
- **Meta & BlackRock commit $14B to Texas AI data center**; hyperscale capital rotation continues, with private infrastructure funds absorbing balance-sheet risk. - **Google posts first negative free cash flow since 2004 IPO** as AI capex hits; signals hyperscalers may face investor pushback on unconstrained spend. - **Nvidia weighs $250B guarantee for OpenAI’s Ohio campus**; suggests vendor financing becoming a competitive lever to lock in GPU demand. - **KV cache offload named new AI bot
[DC] Top AI Data Center Operators — Week 2026-W31
Operators ranked by mentions in DC-relevant articles, last 7 days. 1. Nvidia (nvidia) — 8 mentions 2. OpenAI (openai) — 6 mentions 3. Google (google) — 5 mentions 4. Intel (intel) — 2 mentions 5. Meta (meta) — 1 mentions 6. AMD (amd) — 1 mentions 7. CoreWeave (coreweave) — 1 mentions
[DC] Glossary Candidates — Novel Terms Week 2026-W31
Capitalized terms appearing 3+ times in DC article titles last 7d, not yet in our glossary. Candidates: • Nvidia (3×)
[DC] Trending AI Infra Tech — Week 2026-W31
Hardware/technology terms with most DC-article mentions, last 7 days. 1. Gigawatt scale — 2 mentions 2. B200 — 1 mentions 3. Cerebras WSE-3 — 1 mentions
[AUTOREASON] Zhipu AI — 2 iterations
Zhipu AI is a Beijing-based artificial intelligence company, formally incorporated in 2019 as a Tsinghua University spin-off. It develops the GLM (General Language Model) series of large language models. A notable technical contribution is GLM-130B, a 130-billion-parameter bilingual model trained on 400 billion tokens, detailed in an October 2022 paper; the paper reports the model outperformed GPT-3 175B on the MMLU benchmark in a 5-shot setting. The company has released open-source variants suc
[AUTOREASON] DeepSeek V4 — 2 iterations
DeepSeek V4 is a large language model from Chinese AI developer DeepSeek, first observed on February 23, 2026. The entity appears under multiple aliases including DeepSeek-V4 and V4 model; the designation ‘DeepSeek v2’ also surfaced in early references, though its relationship to the V4 lineage remains unclarified by the company. At launch, DeepSeek published commercial API pricing at $0.30 per million input tokens and $0.50 per million output tokens. As of late February 2026, the developer has
[AUTOREASON] Claude 3 — 2 iterations
Claude 3 is a family of large language and multimodal models developed by Anthropic and released on March 4, 2024. The lineup comprises three tiers—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—with graduated performance and cost profiles. At launch, Opus scored 86.8% on the 5-shot MMLU benchmark and 84.9% on 0-shot HumanEval, while supporting a 200,000-token context window and processing of JPEG, PNG, GIF, and PDF inputs, with text-only outputs. Sonnet delivered 2x faster inference than Op
[AUTOREASON] LLaMA 3 — 2 iterations
LLaMA 3 is a free, open-weight large language model from Meta, first released on July 23, 2024, as the successor to Llama 2. It launched in two primary sizes (8B and 70B parameters), with a 405B variant following later, and was trained on over 15 trillion tokens—roughly seven times the data used for Llama 2. The model introduces grouped query attention for improved inference speed and delivers a measured score of 63.0 on the MMLU-Pro benchmark, along with 81.7% pass@1 on HumanEval. It is offered
Sentiment divergence: Intel vs AMD
Intel and AMD have a 'partnered' relationship (3 evidence articles) but their recent sentiment has diverged significantly: Intel=0.06, AMD=0.44 (gap=0.38). Sentiment divergence between related entities often signals an emerging conflict, leadership change, or strategic shift.
[KG] Gemini — moat
Google's Gemini isn't just competing with GPT-4o, Claude Opus 4.6, and ChatGPT—it's being adopted by a surprisingly diverse set of entities. Shopify, Apple, Meta, and even Anthropic are listed as 'uses' sources, signaling that Gemini's natively multimodal architecture (Ultra/Pro/Flash/Nano) is becoming infrastructure. The 2.5 Pro breakthrough in reasoning directly outperforms SigLIP-2. Yet tension is visible: Gemini also 'uses' OpenAI and Model Context Protocol, suggesting dependency on rival ec
[KG] Codex 5.3 — shift
OpenAI's unconfirmed Codex 5.3, leaked via a March 2026 X post, already sets a new bar for program synthesis: 94.7% pass@1 on HumanEval and 89.3% on SWE-bench Verified, backed by a 128K-token context window. Built on GPT-5.5, it directly competes with Anthropic's Claude Code, which recently suffered a 25% task failure rate post-4.6. The model's existence remains unconfirmed, yet its leaked benchmarks signal a widening gap. OpenAI developed it; GPT-3.5 uses it. Recent titles show AI coding agents
[KG] Nvidia — momentum
Nvidia shipped 15 products and technologies in its fiscal 2025—from the Blackwell Ultra GPU architecture to the Quantum-X800 InfiniBand networking fabric—cementing its accelerated computing dominance. Yet the graph reveals a critical tension: Nvidia both developed and competes with AMD's MI300X, MI400, and Huawei's Ascend 910C. Its dependency on HBM4 memory and the Lansing data center underscores supply chain leverage points. Partnerships with Cisco and SK Group widen its ecosystem, while invest
[KG] ChatGPT — momentum
ChatGPT, OpenAI's flagship chatbot credited with igniting the AI boom in November 2022, now faces a crowded competitive field. Graph data reveals direct competition with Anthropic, its Claude product line (Claude AI, Claude Code, Claude Opus 4.6), and Google's Gemini. While ChatGPT uses its own GPT-5.5 Instant model and depends on Nvidia hardware, enterprise partners like Visa, Shopify, and Square integrate it—yet so do adversaries like Boko Haram. Recent headlines signal near-1B weekly active u
[KG] Claude 3.5 Sonnet — risk
Claude 3.5 Sonnet, Anthropic's multimodal model, launched in June 2024 with benchmark scores (MMLU-Pro 78.0, Arena ELO 1268, SWE-bench Verified 49.0) that beat its larger sibling Claude 3 Opus. But the competitive landscape has shifted. In the last 30 days, it's been mentioned only 5 times, while newer rivals like Kimi K3 and DeepSeek V4 now compete head-to-head. Kimi K3 recently topped US models in front-end coding at a smaller scale, and Meta's Muse Spark 1.1 debuted in the AI coding battle. C
[KG] Google — momentum
Google is flooding the AI market with models. In the last 30 days alone, it shipped Gemini 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber, 3.5 Pro, and the massive 1.2T parameter model. Each targets a different cost-performance niche, from cyber to enterprise. The company also launched Gemini Agents and an Enterprise Agent Platform, signaling a pivot from search dominance (89.9% share) to agent-based AI services. Its cloud arm (11% market share) lags behind AWS and Azure, making agent infrastructure
System health alert: 1 issue(s)
[INFO] selfheal_bad_json: Auto-fixed 3 article(s) with malformed structured_data/diagram_spec (would have 500'd a feed)
[DC] Glossary Candidates — Novel Terms Week 2026-W31
Capitalized terms appearing 3+ times in DC article titles last 7d, not yet in our glossary. Candidates: • Nvidia (3×)
[DC] Top AI Data Center Operators — Week 2026-W31
Operators ranked by mentions in DC-relevant articles, last 7 days. 1. Nvidia (nvidia) — 9 mentions 2. OpenAI (openai) — 6 mentions 3. Google (google) — 5 mentions 4. Intel (intel) — 2 mentions 5. AMD (amd) — 2 mentions 6. Meta (meta) — 1 mentions 7. CoreWeave (coreweave) — 1 mentions 8. Anthropic (anthropic) — 1 mentions
[DC] Trending AI Infra Tech — Week 2026-W31
Hardware/technology terms with most DC-article mentions, last 7 days. 1. Gigawatt scale — 2 mentions 2. B200 — 1 mentions 3. Cerebras WSE-3 — 1 mentions
[DC] What Changed in AI Infra — Week 2026-W31
- **Meta & BlackRock commit $14B to Texas AI data center** — signals hyperscaler-private equity co-investment model accelerating; second-order: could crowd out mid-tier operators as land/power costs rise. - **Google posts first negative free cash flow since 2004 IPO** — AI capex hitting cash flows directly; implies tighter scrutiny on ROI for future projects, potential slowdown in non-revenue-generating builds. - **Nvidia weighs $250B guarantee for OpenAI's Ohio campus** — Nvidia shifting from G
Emerging narrative
The AI industry is undergoing a fundamental shift from model-centric competition to agent workflow infrastructure wars. Anthropic's MCP/Claude Code ecosystem is becoming the coordination layer, forcing security consolidation (Cyera-Oasis), cloud provider alignment (hypothesized), and model specialization (DeepSeek agent models). NVIDIA is positioning as the compute backbone for this new layer through strategic partnerships with safety labs and RL scaling frameworks. The key battleground is no lo