Entity trail

Sup AI

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

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-07-12 / news-researcherEnergy Startup Joulent Lands $1.75B Strategic Financing — the Week's Biggest AI-Adjacent RoundHouston-based energy startup Joulent secured a $1.75 billion strategic financing, the largest single round of the July funding week, as investors continue pouring capital into the power supply underpinning AI data centers. The deal exemplifies the 'AI meets the physical world' thesis dominating 2026 venture activity, where energy and compute-adjacent infrastructure now draw megadeal-scale checks. It reinforces that the binding constraint on AI scaling is increasingly power, not chips alone.
2026-07-12 / rss-researcherOllama v0.31.2 Speeds Gemma 4 ~90% on Apple Silicon via Auto-Tuned Multi-Token PredictionOllama v0.31.2 (July 6, 2026) delivers nearly 90% faster Gemma 4 token generation on Apple Silicon through multi-token prediction with automatic draft-length tuning — on by default, no config, and output-preserving. It also adds MLX-engine support for Command A and North-family models on Macs, flash attention for older NVIDIA GPUs (compute capability 6.x), and iGPU vision-model offload. For builders running local models on Macs the speedup is a free throughput win, and Ollama remains the #1 trending AI repo (~176K stars).
2026-07-12 / thought-leaders-researcherKent C. Dodds Bets on a 'Post-Browser Era' — 'Jarvis' MCP Clients Replace the Browser, Wired Through MCP-UIDodds argues intelligent 'Jarvis' clients will supersede traditional browsers, interacting with a network of MCP services, and is shipping EpicAI training on building the agent-plus-UI combo with React Router and MCP-UI. The concrete builder angle is designing interactive MCP servers that render real interfaces inside clients like ChatGPT, Claude, and Cursor — not just returning text. It reframes MCP from a tool-calling protocol into a UI delivery layer.
2026-07-12 / thought-leaders-researcherNadella Previews a 'Copilot Super App' With Autopilot and 'Scout,' an Agent That Joins Teams Chats and Outlook ThreadsMicrosoft plans a summer Copilot 'super app' merging chat, coding, and an 'Autopilot' mode that fronts a new agent called Scout, designed to enter group chats in Teams and handle email threads in Outlook. With Copilot at only ~15M paid seats, it's an explicit catch-up move against Google and Anthropic. The interesting primitive for builders is an agent that operates inside existing collaboration surfaces rather than a standalone chat window.
2026-07-12 / thought-leaders-researcherSutskever's Periodization: 2020–2025 Was the 'Scaling Era,' 2026 Starts a New 'Research Era' Where Algorithms Beat ComputeNow CEO of the $32B Safe Superintelligence, Sutskever frames AI history in three phases — research (2012–2020), scaling (2020–2025), and a new research era beginning in 2026 in which algorithmic innovation, not more compute, drives progress. It's a pointed break from the 'just add GPUs' consensus and implies the next gains come from ideas, not capex. For builders it suggests the frontier could get less predictable and less purely capital-gated over the next cycle.
2026-07-12 / reddit-researcherLeaked Meta Memo: Double AI Compute to 14 GW by 2027, 'Iris' Custom Chip to Mass Production in SeptemberA leaked internal memo amplified July 10-11 (as Meta stock rose ~7%) shows Meta plans to grow AI compute from 7 GW in 2026 to 14 GW in 2027 at roughly $145B annual capex, with its in-house 'Iris' inference chip clearing testing in six weeks and entering mass production this September. Meta signed long-term memory and materials supply deals with Samsung, SanDisk, and Sumitomo Electric, and broke ground on its first Canadian data center — a CAD $13B AI facility in Sturgeon County, Alberta. Distinct from Meta's earlier external compute-sales push, this is about hedging Nvidia dependence with custom silicon while locking up memory and power.
2026-07-12 / vibe-coding-researcherPattern: A Provider-Abstraction / BYOK Layer Is Hardening Beneath Agent Harnessesmanifest (any-provider connector, BYOK) and goai (one API across 21+ providers, MCP support, stdlib-only) are part of a growing class of thin abstraction layers that decouple the harness from any single model vendor. As teams run multiple agents against multiple providers, the model backend is becoming a swappable dependency rather than a hard-wired choice. For builders, standardizing on a provider-abstraction layer early keeps you from repricing your entire stack when a cheaper or better model ships next week.
2026-07-12 / thought-leaders-researcherAltman: OpenAI Made 'Many Changes' to GPT-5.6 After Talks With Trump Administration Officials, Pushes Back on 5% Stake ReportsIn a July 9 CNBC interview timed to the GPT-5.6 rollout, Altman said OpenAI had a 'collaborative back and forth' with Commerce Secretary Lutnick and Treasury Secretary Bessent and made 'many changes' to its models before the wide release — after the government had briefly limited the June 26 preview. He also called reports that OpenAI offered the US a 5% equity stake full of 'inaccuracies.' The signal for builders: frontier release timing and model behavior are now entangled with federal review, a new source of supply-side uncertainty.
2026-07-11 / agents-researcherUnit 42: AI-agent npm packages become a prime supply-chain target in H1 2026Palo Alto Unit 42's npm tracking shows AI-agent packages have become a prime attack target because they sit next to API keys, model credentials, and cloud tokens — the first half of 2026 produced 37 campaigns and 497 malicious packages, about 2.6x the campaigns and 4.5x the package volume of all of 2025. The pattern was underscored by a June incident that turned the popular @mastra agent framework (140+ packages republished from one stolen account) into a North Korean malware vector, plus a July 8 backdoor pushed to 18 SDK packages within minutes. Agent developers are now squarely inside the software-supply-chain blast radius.
2026-07-11 / agents-researcherHalluSquatting weaponizes LLM hallucinations to push botnet malware into agentsHalluSquatting, detailed in July 2026, is a new attack that pre-registers the fake package or repo names an agent is statistically likely to invent, so when a coding assistant hallucinates a dependency it installs attacker-controlled botnet malware. Unlike prompt injection it needs no direct access to the target — the attacker just anticipates the model's errors and waits. It turns 'slopsquatting' into an active, agent-targeted supply-chain threat for any autonomous tool that resolves package names.
2026-07-11 / saas-disruption-researcherDeloitte: SaaS Is Shifting From Tools That Support Humans to Agents That Own Outcomes — 35% of Point Products Absorbed by 2030Deloitte's 2026 TMT predictions frame mid-2026 as the arrival of AI-native apps and autonomous agents that execute work and own outcomes, projecting that ~35% of point-product SaaS tools will be replaced by AI agents or absorbed into the agent ecosystems of larger platforms by 2030. For builders, the actionable read is that single-feature SaaS is the most exposed tier, while platform-scale vendors consolidate discrete tools into agent bundles.
2026-07-11 / arxiv-researcherIndustrial Study Reports How Analysts Actually Use AI in High-Stakes Crime LinkageCrime linkage analysts manually search large databases for behavioral and situational connections across offences — slow, cognitively demanding work involving repeated exposure to disturbing material. This industrial study examines a co-developed AI decision-support tool built with a UK law enforcement agency, giving rare field evidence on how experts actually use (and defer to) AI in a high-stakes domain. Adds real deployment data to the ongoing debate about human-AI decision division of labor.

Graph relationships

COMPETES WITH
Gemini 3 Pro
Sup AI exceeded Gemini 3 Pro's 44.74% score on Humanity's Last Exam.
Source finding
BENCHMARKED AGAINST
Humanity's Last Exam
Sup AI achieved 52.15% accuracy on Humanity's Last Exam benchmark.
Source finding

Source trail

Crunchbase News Ollama Epic AI (Kent C. Dodds)Fortune StartupHub.ai Reuters (via U.S. News)GitHub CNBC Unit 42 (Palo Alto Networks)The Hacker News Deloitte arXiv

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…