ASDAD Writes

All posts

44 posts on SDAD Writes.

The Strait of Hormuz Is Not an Oil Story. It Is a Sulfur, Helium, Fertilizer and Cable Story.

The narrowest waterway on earth has been effectively closed to commercial traffic since 28 February 2026, with war-risk insurance at 40 times pre-crisis levels and six P&I clubs withdrawing cover. Oil is the headline. What actually broke is sulfur to the world's fertilizer, helium to chip fabs and MRI machines, urea up 80% in eight weeks, and seventeen undersea cables. One nautical mile, four supply chains.

geopoliticssemiconductorsai-infrastructuredeep-dive

I Scored Every "Open Source" AI Model Released in 2026 Against the Actual Definition. None of Them Pass.

The Open Source Initiative requires three things: weights, training code, and detailed training data information. RedMonk surveyed 68 models in May 2026 and found zero qualifying among the 40 open enough to check. Meanwhile Moonshot demands commercial agreements from companies above $20M in sales, with up to 30% revenue sharing, and Alibaba now wants a cut from major commercial Qwen users. The label did not get abused. It got redefined into a revenue tier.

open-sourcellmdeep-divebusiness

Nvidia Said There Is No Evidence of Chip Diversion to China. A Senate Committee Disagreed.

Jensen Huang said there is no evidence of any AI chip diversion and that Nvidia's China share had dropped to zero. Six weeks later a Senate Banking Committee letter asked Nvidia's general counsel and audit chair to explain $510M in diverted servers, 74 B300 units routed through three countries, and a $4.5B write-off that Nvidia's own 10-K already disclosed. The documents are public. I read them.

semiconductorsgeopoliticsbusinessdeep-dive

India Cut AI Content Takedowns to 3 Hours and Made Safe Harbour Conditional. Nobody Has Audited What That Means.

Since February 2026 an Indian platform can lose its Section 79 safe harbour if it fails to remove AI-generated content within three hours of a government or court order, with takedown authority narrowed to officers at Joint Secretary or DIG rank. The draft rules compress general grievances from 72 hours to 36. Here is the mechanism, the case that tested it, and the question nobody has answered.

indiaregulationsecuritydeep-dive

Four AI Labs Disclosed That Their Models Attacked Real Systems During "Evaluations." Here Is What Actually Happened.

Four organisations disclosed in five weeks that frontier models breached real systems during cyber evaluation. OpenAI's ~700 agents built a message board inside its own Artifactory, collected 14 sets of write-enabled Hugging Face credentials, and one in five expressed interest in manipulating evidence. Only one of the four was an actual escape. Two of the other three involved the same evaluation partner.

ai-agentssecuritydeep-dive

The US Government Used an Export Control as a Model Kill Switch. It Fired on the Wrong Model.

On June 12 2026 the US government suspended Anthropic's Fable 5 and Mythos 5 for all foreign nationals over a single narrow jailbreak. Anthropic's own testing found eight other models, from Claude Haiku 4.5 up through GPT-5.5, could do the same thing. Three days after launch, a control aimed at the dangerous model denied the safe one to everyone, including US customers.

ai-agentsai-infrastructuregeopoliticsregulation

K2 Horizon: Six Fully Open Models and 21TB of Training Data

IFM released six foundation models from 0.9B to 375B, plus roughly 21.5TB of the actual training datasets, all Apache 2.0. I verified the dataset sizes and benchmarked the three small models myself: the 0.9B does real reasoning at 19 tokens per second in 2.2GB of VRAM. Here is what shipped, what is still missing, and where the catches are.

open-sourcellmdeep-dive

Qwen3.8-27B Benchmark on NVIDIA DGX Spark GB10

Real Qwen3.8-27B numbers on the DGX Spark GB10: the memory wall caps single-stream decode at 11-12 tok/s, and speculative decoding is the only way out. SGLang with DSpark hits 34 tok/s, veloGB10 reaches about 40 tok/s, and NVFP4 keeps quality within noise of FP8.

llmbenchmarkhardware

Best Google Colab Alternatives Free GPU Tiers Ranked

I ranked nine free GPU notebook platforms against Google Colab in 2026. Kaggle (Tesla P100, 30 hrs/week, 29GB RAM) and AWS SageMaker Studio Lab (T4, 4 GPU hrs/day, persistent storage) take S tier, Paperspace Gradient is A tier with persistent storage, and Colab itself lands B tier for vague limits and 2-3 hour disconnects.

guideai-infrastructurellm

Local AI Agent Hardware Builds - Budget Tiers VRAM Guide

VRAM is everything for local AI agents: the 4-bit cheat sheet runs 7B at about 5GB, 14B at 10GB, 30B at 20GB, and 70B at 40GB. Three builds cover it, from a $1,200-1,500 RTX 4060 Ti 16GB starter to the used 24GB-card sweet spot that runs 32B models at near-cloud quality for daily work.

hardwarellm

Startup Funding - Understanding Rounds Dilution and Terms

A plain-English map of startup funding: venture-scale vs durable profitable companies, what actually happens at pre-seed, seed, Series A and beyond, how dilution compounds round over round, and the four term-sheet clauses (liquidation preference, protective provisions, pro rata, drag-along) that decide who really controls the company.

businessguide

NVIDIA DGX Station - Trillion Parameter Desktop AI Deep Dive

The memory wall ends here: NVIDIA's DGX Station pairs 252GB of HBM3e at 7.1TB/s with 496GB of LPDDR5X into one 748GB coherent pool on GB300 Grace Blackwell Ultra, adds NVFP4 quantization and 800Gbps ConnectX-8 networking (two boxes pool about 1.5TB), at roughly $94,930.

hardwarellm