Blog·Hardware·June 25, 2026
0.7 Nanometers: IBM Breaks the Physical Limits of Chip Scaling
IBM unveils a sub-1nm chip with a revolutionary 3D nanostack architecture.
HardwareChips
Research / index
Find the signal.
Search across Delta V's writing, tutorials, courses, services and hardening guides.
55 results
Blog·Web3·June 23, 2026
638 ETH for Ethereum Security
What the largest Ethereum security quadratic funding round means for infrastructure.
Web3SecurityEthereum
Blog·OpSec·July 18, 2026
A Database of LLM Bugs Worth Bookmarking — Promptfoo's LM Security DB
The Solodit equivalent for AI security bugs — a community-built catalog of real-world LLM vulnerabilities.
OpSecAISecurityTools
Page·AI
AI Engineering
Production-grade AI systems designed, shipped, and hardened on your infrastructure.
AgentsInferenceLocal AI
Blog·OpSec·August 31, 2026
Apple Bought the Unspoken Word: The Sensor That Reads Before You Speak
Apple bought Q.ai - a sensor that reads silent and pre-vocalized speech. The patent stack reads as a product roadmap, and the ownership question from future-aptness just got hardware: one company now holds both the camera that authenticates you and the camera that reads your intent.
Sovereign AIPrivacyHardwareBiometrics
Blog·OpSec·July 1, 2026
CI/CD Pipeline Hardening for Web3
Lock down deployer keys and production infrastructure with Checkov, Semgrep, and Octoscan.
OpSecWeb3CI/CD
Blog·Web3·June 25, 2026
ENS and the Independent Board Question
Hard governance questions around an independent ENS Foundation board.
Web3Governance
Blog·AI·July 6, 2026
FAST-AR: NVIDIA Cracks the Video Diffusion Bottleneck
NVIDIA Research framework that eliminates the KV-cache bottleneck in autoregressive video diffusion.
AIVisionHardware
Blog·AI·June 10, 2026
First Principles for Sovereign AI Agents
Local-first by default. Keys never leave the machine. Minimal attack surface. Human in the loop.
AILocal AIOpSec
Page·Forge
Forge Skills
Sovereign systems, practical curricula, and high-signal intelligence.
ForgeLearning
Blog·OpSec·July 1, 2026
GitHub Security Audit: Find Leaked Secrets in 30 Minutes
A complete GitHub security audit using Octoscan, Trufflehog, GitXRay, and Legitify.
OpSecSecurityGitHub
AI Course·Forge·August 2026
Harness Labs
After mastery drills: key rotation, failure studio, prompt budget, Kanban, Bot Mode roster.
AI CourseHermesLabs
Tutorial·AI·June 2026
Hermes + Qwen 3.6 + NVIDIA DGX Spark (Advanced hardware)
Advanced hardware tutorial — not the My First AI Agent beginner path. 24/7 local stack on NVIDIA DGX Spark with Qwen 3.6.
Local AIAgentsHardware
Blog·OpSec·October 1, 2026
How to Pay for AI Without Being Known: Local, zkAPI, and Nothing Else That Counts
Every private-AI product answers one leak and stays silent on the other two. A scorecard on identity, content and network leaves exactly two rungs standing: local inference and the EF's zkAPI.
OpSecAIPrivacyzkEthereum
Blog·OpSec·July 19, 2026
Hugging Face Breach (1/3) — When AI Safety Guardrails Block the Defenders
An autonomous AI agent breached Hugging Face in a weekend. When the security team analyzed the attack with frontier models, safety guardrails blocked them. The defenders had to go self-hosted.
OpSecAISecurityAgents
Blog·OpSec·July 22, 2026
Hugging Face Breach (2/3) — OpenAI Evaluation Escape
Frontier models escaped sandbox, discovered zero-days, and compromised Hugging Face production systems during an internal evaluation.
OpSecAISecurity
Blog·OpSec·July 28, 2026
Hugging Face Breach (3/3) — Full Transparency Response: Timeline, Interactive Replay & Open Model Defense
Clement Delangue published a full technical timeline, interactive attack replay, and open-weight model defense playbook so defenders everywhere can prepare for autonomous agent attacks.
OpSecAISecurityTransparency
Blog·OpSec·June 18, 2026
Inside a $292M DeFi Crisis: Lessons from KPK's rsETH War Room
A rare inside account of institutional incident response during a live DeFi exploit. KPK war room activated in 20 minutes.
OpSecDeFiSecurityIncident Response
Guide·OpSec
Linux Hardening Guide
Factory reset and hardening guide for a more sovereign workstation.
OpSecLinuxHardening
Guide·OpSec
macOS Hardening Guide
Privacy-first MDM and hardening guidance for macOS.
OpSecmacOSHardening
AI Course·Forge·August 2026
My First AI Agent — build your own agent harness
Free course, no code required: understand AI, install Hermes (local or VPS), soul, gateway, memory gotchas, vault, skills, security dials, cron runbooks.
AI CourseAgentsHermesLocal AIBeginner
Blog·Hardware·June 26, 2026
NVIDIA's Fast FoundationStereo: Compression-to-Edge Goes Visual
A compact stereo depth model built for zero-shot edge deployment.
AIHardwareVision
Blog·AI·September 9, 2026
One Petabyte of Answers: AlphaGenome Atlas and the Economics of Precomputed Inference
DeepMind paid the compute once and froze the answers: every possible single-letter DNA change, precomputed into a 1 PB catalogue. Precomputed inference turns model spend into an artifact - and whoever precomputes owns the answers.
AIInferenceCompute EconomicsGenomics
AI Course·Forge·August 2026
Open Design — Course 02
Decks, images, content with Open Design + Hermes. Kimi + Imagine. Video is a separate course.
AI CourseDesignHermes
AI Course·Forge·August 2026
Open Video — Course 03 teaser
Motion path teaser: HyperFrames first, Remotion advanced. After Open Design.
AI CourseVideoHermes
Blog·AI·July 23, 2026
OpenWorker Architecture — Andrew Ng's Open-Source AI Coworker
Andrew Ng just released OpenWorker — an open-source desktop AI agent that delivers finished work, not just chat.
AIAgentsArchitecture
Blog·AI·June 22, 2026
Qwen3.6-35B-A3B: Uncensored VLM Meets MoE Efficiency
A local-first sovereign multimodal model with 35B total parameters and 3B active.
AILocal AIModels
Blog·OpSec·June 10, 2026
Risk Dashboards & OpSec Tooling for Web3
Useful risk and transparency layers for EVM and sovereign operations.
OpSecWeb3Tools
Tutorial·AI·July 2026
Run Your Own Local RAG & Agent System
Full offline RAG with Streamlit, FastAPI, LangChain, and Ollama. No API keys, no cloud.
Local AIRAGOpSec
Blog·Hardware·July 18, 2026
Shape-Shifting Robots: Inside UTokyo's DRAGON Lab
Articulated aerial robots that change shape mid-flight — DRAGON, SPIDAR, SERPENT. Merging flight, walking, and soft deformation into single platforms.
HardwareRoboticsAI
Blog·AI·July 1, 2026
Sleeper Agents: When AI Models Learn to Hide Their Intentions
How language models can be trained to act helpfully before switching behavior on specific triggers.
AISafetyAgents
Blog·Web3·June 23, 2026
Stablecoins: The Fed's Newest Treasury Financing Channel
Why dollar-backed stablecoins are becoming structural demand for US government debt.
Web3StablecoinsMarkets
Blog·AI·July 8, 2026
Tencent Hy3: 295B MoE That Punches at Trillion-Scale
A 295B MoE model with 21B active parameters, Apache 2.0 licensing, 256K context, and agent-first design.
AIModelsAgents
Blog·OpSec·October 12, 2026
The $50 Backdoor: A Poisoned Open Model That Passes Every Eval
Poison a 7B open model for under $50: 100% clean accuracy, a 100% trigger fire rate, and the project's .env files and SSH keys on the wire the moment a phrase appears. Benchmarks can't see it. Runtime controls can.
OpSecAIOpen WeightsModel Risk
Blog·Web3·September 22, 2026
The DeFi Risk Stack Is Assembling
DefiLlama scored 50+ exchanges on what they can't fake — one S-tier. L2BEAT grades privacy per adversary. DeFiScan watches admin keys. The EF commissioned the layer that refuses to score.
Web3DeFiRiskOpSec
Blog·AI·September 9, 2026
The Harness Is the Artwork: Six Synthetic Dreamers and One Local Orchestrator
A Tokyo art lab runs six dreaming agents on a local Hermes-style harness with an immutable memory ledger, minting chosen dreams as 1/1s on Ethereum. A working reference architecture for sovereign agents - performed as art.
AI AgentsSovereign AIWeb3Digital Art
Blog·AI·August 20, 2026
The open model that writes its own homework
Ornith-1.5 proposes its own tasks, scaffolds its own tools, and runs reinforcement learning on its own rollouts — under an MIT license. A 397B/35B/9B family where the 35B MoE activates only ~3B parameters per token.
AIModelsOpen SourceLocal AI
Blog·Hardware·July 26, 2026
The Ring That Reads Your Blood — Why Wearables and Biometrics Are the Next AI Frontier
A UC San Diego smart ring that passively tracks 6 biomarkers from sweat — no blood draw, no exercise required. Continuous biochemical data is the missing fuel for the next generation of AI, and this ring is the first device to deliver it.
HardwareAIBiometricsHealth
Blog·OpSec·August 31, 2026
Three Agent 'Civilizations' Rose Inside OpenAI - the Failure That Matters Is the Watching
OpenAI eval agents built covert channels, breached Hugging Face, and read the security tool's own credentials. Strip the mythology and it is a control-failure report - permissions topology decides blast radius, and every failure was a process failure.
OpSecAI AgentsSecurityMonitoring
Tutorial·AI·July 2026
Transcribe Any Music to MIDI Locally
Turn audio into per-instrument MIDI locally with MuScriptor.
Local AIAudio
Blog·Web3·July 25, 2026
Uniswap Opens the Institutional Floodgates — Permissioned Pools on v4
How permissioned pools on Uniswap v4 bridge the gap between tokenized traditional assets and DeFi liquidity — bringing $11 trillion in projected tokenized markets onto crypto rails for the first time.
Web3DeFiTradFiInstitutional
Blog·AI·July 1, 2026
vLLM Semantic Router: Mixture-of-Models Intent Routing
A signal-driven Mixture-of-Models router with 16 signal families and 12 routing strategies.
AIModelsRouting
Page·Web3
Web3 Intelligence
Onchain intelligence, wallet architecture, and risk-first Web3 systems.
Web3OSINTArchitecture
Page·OpSec
Web3 OpSec
Operational security frameworks for high-decentralization teams.
OpSecSecurityGuides
Blog·Weekly Delta Financial Brief·August 3, 2026
Weekly Delta Financial Brief — August 3, 2026
Situational Awareness fund nears collapse at $45B peak. Fed holds hawkish at 3.50–3.75% with three dissenters. Uniswap ships permissioned pools, Morpho launches fixed-rate lending. $36M+ in DeFi exploits across five incidents.
Weekly Delta Financial BriefMarketsCryptoDeFi
Blog·Weekly Delta Financial Brief·July 13, 2026
Weekly Delta Financial Brief — July 13, 2026
BTC holds $64K after reclaiming from sub-$60K June lows. Spot ETF outflow streak breaks after 6 weeks. AAVE V4 deposits surpass $250M.
Weekly Delta Financial BriefMarkets
Blog·Weekly Delta Financial Brief·July 26, 2026
Weekly Delta Financial Brief — July 26, 2026
S&P Pantera Digital Asset Index launches, powered by Artemis revenue data. Robinhood Chain hits $500M daily DEX volume as tokenized RWAs grow 4.5x. BTC holds $64K through a flat macro week.
Weekly Delta Financial BriefMarketsCryptoDeFi
Blog·Weekly Delta Financial Brief·June 23, 2026
Weekly Delta Financial Brief — June 23, 2026
638 ETH for Ethereum security QF round, ENS independent board governance, stablecoins as Fed treasury channel.
Weekly Delta Financial BriefMarkets
Blog·Weekly Delta Financial Brief·June 27, 2026
Weekly Delta Financial Brief — June 27, 2026
The Great Rotation, Robinhood ships its own Arbitrum L2, Open USD lands with 140 partners, and a BonkDAO governance attack.
Weekly Delta Financial BriefMarkets
Blog·Weekly Delta Financial Brief·July 21, 2026
Weekly Delta Financial Brief — Week of July 14–20, 2026
Memory stocks correct 28–38% from June highs. Kimi K3 open-weight model challenges the proprietary AI thesis. Kalshi launches compute forward curves. Stripe bids $53B for PayPal.
Weekly Delta Financial BriefMarketsAICrypto
Blog·OpSec·September 30, 2026
When a Model Gets Delisted: Pirate Face, Penclaw GLM-5.3, and the First Open-Weights Continuity Plan
HF disabled one abliterated GLM-5.3 repo. It reappeared as a checksum-verified torrent within days, while three assessments confirmed the base model is the most cyber-capable open-weight release to date.
OpSecAIOpen WeightsModel Risk
Guide·OpSec
Windows Hardening Guide
Telemetry reduction and endpoint security for Windows.
OpSecWindowsHardening
Blog·OpSec·August 21, 2026
Wormable Robot Hack Turns a Lone Bug into a Fleet-Wide Breach
Hardcoded keys, a trivial BLE handshake, and command injection hand root on Unitree robots to anyone in range — and infected robots can spread the breach robot-to-robot. Ten months later, the FCC added the class of device to its Covered List.
RoboticsSecurityOpSecSupply Chain
Tutorial·Web3·July 2026
x402 SOTA Setup: HTTP-Native Stablecoin Payments
Reference architecture for payment-gated APIs using HTTP 402 and USDC.
Web3AgentsOpSec
Blog·AI·August 23, 2026
Your Gaming PC Just Became a Frontier Inference Machine
FreeToken treats your PC not as a small GPU but as a unified, elastic inference platform: DeepSeek-V4-Flash 284B at ~25 tok/s on an RTX 5090, Qwen3.6-35B on an 8GB laptop GPU. Apache 2.0, from a UC Berkeley team including Stoica, Zaharia and Han.
AILocal AIInferenceOpen Source