open-models — tisram

The Wall Street Journal 2026-10-27-2

AI Hacks Are Coming Thick and Fast and Overwhelming the Internet

Parsons was billed $449 across about 50 charge emails overnight after attackers drained Anthropic and OpenAI credentials stored in OpenCode, and other users reported $1,700. The target was inference spend, so a stored key behaves like a card number. Governments in Canada, Belgium, France and the Netherlands took sites offline rather than wait for patches against bugs that could expose up to 70% of websites, the move Sima's account says stops working once the attacker holds your own keys.

# tags

ai-cybersecurity ai-security agent-security vulnerability-management patch-velocity incident-management open-models kimi opencode crowdstrike wsj

◆ entities

Risk management Open Source Strategy Blast Radius Agent governance Enterprise AI security posture

→ threads

ai-cybersecurity ai-security

⟷ links

2026-04-01-3 2026-04-13-3 2026-05-22-w2 2026-04-01-1 2026-05-14-1 2026-04-05-1 2026-05-22-3 2026-04-04-2 2026-03-13-w3 2026-05-12-2

permalink

Foreign Affairs 2026-10-22-2

Empires of AI: The Rise and Fall of a New Global Order

The Empires of AI piece argues that scaling laws tie the frontier to energy, chips and capital, so capital decides who can keep racing. The US paid about $881B in interest against $874B on defense in 2024, while the biggest tech firms committed more than $600B of 2026 capex. Add the June Commerce order that took Anthropic's top two models dark worldwide, and access to the output of those scarce tools looks like a revocable state concession.

# tags

ai-geopolitics ai-capital-cycle export-controls china-ai sovereign-ai taiwan-risk guns-and-butter historical-analogy world-order open-models scaling-laws national-security compute-scarcity tsmc foreign-affairs

◆ entities

National security Fiscal dominance AI capex cycle Geopolitical Risk Supply Constraint Economics

→ threads

ai-capital-cycle compute-scarcity

⟷ links

2026-03-13-w3 2026-04-12-1 2026-04-03-2 2026-05-18-1 2026-03-18-1 2026-05-09-w1 2026-10-21-3 2026-03-11-2 2026-03-18-3 2026-10-16-1 2026-10-16-w2

permalink

In Good Company with Nicolai Tangen 2026-10-21-2

Eric Schmidt: The Age of Agents, Superintelligence, and Why Europe Is Falling Behind

Schmidt says the gains are real but concentrated in software, cyber and math, where learning loops are fast, and biology lags on data. The effect he is most confident about is organizational: a friend's firm cut 20% of staff, the 'glue' planners, and an estimated 2,000 of 10,000 people at a large firm are unneeded. Read against deBoer's cancer example, displacement arrives first where the work is coordination, not where the problem is physical.

# tags

in-good-company podcast agentic-ai org-design change-management european-ai europe ai-regulation china-ai distillation open-models ai-geopolitics recursive-self-improvement superintelligence workforce-dynamics ai-bubble sovereign-wealth-funds

◆ entities

Recursive self-improvement Teacher-student distillation Organizational absorption capacity Superintelligence AI regulatory risk

→ threads

ai-regulation ai-bubble

⟷ links

2026-05-27-2 2026-04-03-2 2026-03-13-w3 2026-04-10-1 2026-03-08-1 2026-04-12-3 2026-04-05-1 2026-05-01-3 2026-03-30-3 2026-04-17-w3

permalink

VentureBeat · 2026-10-15 2026-10-16-w1

NVIDIA's new, open Nemotron 3.5 Super VL with post-training beats Claude Opus 5, GPT-6.1 Sol on specialized task benchmarks

Super VL, a 120B open base, went from about 28% to about 72% on medical claims after post-training, against about 38% for Claude Opus 5. The base model sat below Opus 5, so the gain came from the customer's data and tuning budget, and the comparison is tuned against untuned. Its Artificial Analysis index of 26 trails GLM-5.3-Flash (42) and Qwen3.8-Flash-Next (40), which says the weights are not what wins. For builders the buying decision moves from which model to who owns the evaluation set and the post-training pipeline. In the week's argument this is the lightest of Nvidia's levers: it gives away the part that is commoditizing and keeps buyers building on its stack, with Bolt E4B due Oct 23 as the next test.

# tags

benchmark edge-ai enterprise-ai fine-tuning hugging-face huggingface inference-economics local-llm model-licensing nvidia open-models post-training vendor-claims venturebeat

BG2Pod · 2026-10-16 2026-10-16-w2

BG2Pod with Brad Gerstner and Bill Gurley — Jensen Huang | Does AI Revenue Justify Capex, '27 AI Bottlenecks, Singularity, Open vs Closed & more

Jensen's claim that the moat is growing rests on a buyer's logic: a $50-100B order can't be benchmarked, so it gets underwritten on fungibility, durability and rentability. Nvidia's content per gigawatt has gone from about $18B to 40-plus B, and about a third of the business is startups renting compute, so payback does not hinge on two labs reaching trillion-dollar revenue. SemiAnalysis expects 43 GW of new AI compute next year, against a Street view of 35-40 GW of land, power and shell that may come online. That is the lender's checklist, and it explains why a guarantee book exists at all. What it adds to the week is the demand side of the financing: the AI-bubble question now sits with the non-lab two-thirds and how fast power gets energized.

# tags

ai-capex ai-capital-cycle ai-infrastructure-capex ai-safety compute-scarcity data-centers jensen-huang moat nvidia open-models orbital-compute physical-ai podcast superintelligence

BG2Pod 2026-10-16-1

BG2Pod with Brad Gerstner and Bill Gurley — Jensen Huang | Does AI Revenue Justify Capex, '27 AI Bottlenecks, Singularity, Open vs Closed & more

Jensen argues the moat is growing because a $50-100B order can't be benchmarked, so buyers underwrite it on fungibility, durability and rentability. Nvidia's content per gigawatt has climbed from about $18B to 40-plus B, and roughly a third of the business is startups renting compute. That is the lender's checklist, and it explains why Hoffman's guarantee book exists at all.

# tags

nvidia jensen-huang ai-capex ai-capital-cycle ai-infrastructure-capex compute-scarcity physical-ai open-models ai-safety orbital-compute podcast superintelligence data-centers moat

◆ entities

AI capex cycle AI safety Moat (competitive advantage)AI Infrastructure Investment Supply Constraint Economics

→ threads

ai-capital-cycle ai-infrastructure-capex compute-scarcity physical-ai ai-safety

⟷ links

2026-03-20-w1 2026-04-21-2 2026-03-31-m2 2026-04-17-w3 2026-05-15-w3 2026-05-01-w2 2026-05-26-2

permalink

Meta AI Research 2026-10-15-1

Introducing Muse Spark Open, The Most Capable US Open‑Weight Model to Date

Meta's Muse Spark Open post (816B total, 42B active, 1M context) is far more specific about availability than capability: seven API hosts, four weight hubs and seven-plus engines on day one, none of them Meta. The local recipe is two 512GB M5 Ultra Mac Studios running through MLX. Meta is giving the weights away to buy presence across every stack, which fits the Spyglass reading that its open tier trails its closed one.

# tags

meta open-models apple-silicon mlx local-llm on-device inference-economics vendor-claims model-licensing ai-capital-cycle blog-post

◆ entities

Open Source Strategy Hardware fragmentation Distribution moat Inference cost economics

→ threads

inference-economics ai-capital-cycle

⟷ links

2026-03-14-3 2026-05-15-w3 2026-04-08-3 2026-04-14-1 2026-03-31-m2 2026-03-20-w1 2026-04-04-3 2026-03-13-1 2026-03-21-3

permalink

CNBC 2026-10-15-2

Apple and Meta team up to make Muse AI model run on Mac computers

CNBC's account of the Apple tie-up supplies the hardware half. Muse Spark Open needs 2 to 4 paired Mac Studios, and an M5 Ultra with 256GB starts at $10,799. Alexandr Wang says the open model is meant to pull developers in small and then scale them up, and by his framing memory capacity, not compute, gates local use. That makes sovereign coding a fixed-capex purchase and puts Apple, not NVIDIA, on the default local path for the largest open models.

# tags

cnbc meta apple apple-silicon open-models local-llm on-device edge-ai mlx ai-infrastructure inference-economics hardware-fragmentation open-source ai-strategy

◆ entities

Open Source Strategy Vendor Lock-in Risk Hardware fragmentation Platformization Inference cost economics

→ threads

inference-economics

⟷ links

2026-04-08-3 2026-04-25-2 2026-04-21-2 2026-04-04-3 2026-03-14-3 2026-03-30-3

permalink

VentureBeat 2026-10-15-3

NVIDIA's new, open Nemotron 3.5 Super VL with post-training beats Claude Opus 5, GPT-6.1 Sol on specialized task benchmarks

NVIDIA's Super VL, a 120B open base, went from about 28% to about 72% on medical claims after post-training, against about 38% for Claude Opus 5. The base model sat below Opus 5, so the gain came from the customer's data and tuning budget, and the comparison is tuned against untuned. Set beside Meta's Muse Spark Open, it suggests the weights are the commodity and the evaluation set and training pipeline are what a buyer should own.

# tags

nvidia open-models post-training fine-tuning enterprise-ai vendor-claims benchmark venturebeat model-licensing hugging-face huggingface edge-ai local-llm inference-economics

◆ entities

Teacher-student distillation Open Source Strategy Ecosystem lock-in through integration breadth Vertical Integration Enterprise AI adoption

→ threads

inference-economics

⟷ links

2026-03-17-3 2026-03-13-1 2026-03-15-1 2026-03-14-3 2026-03-10-1 2026-05-27-3 2026-04-17-w2 2026-03-17-2 2026-05-07-1 2026-03-14-1

permalink

The Deep View 2026-05-07-1

OpenAI MRC Protocol: What Gets Open-Sourced Is the Non-Moat

What frontier labs open-source is a map of the non-moats. OpenAI released its GPU networking protocol through OCP with Microsoft, AMD, Broadcom, NVIDIA, and Intel as coalition partners, two years in development, already running at Stargate's Abilene site and used to train GPT-5.5. The corollary lands hardest for Microsoft: they have the protocol, run it on Fairwater, and still ship mid-class models, which means networking efficiency was never the binding constraint.