feed
last scrape: 3h ago · 961 in archive · 9 today
this week
Interview with geneticist David Reich (transcript) on his new paper showing natural selection accelerated over the last 10k years—+1 SD rise in cognitive performance predictors, mostly Bronze Age—plus his hypothesis that Neanderthals were modern humans genetically swamped by European archaics after a Caucasus expansion.
Reich: Bronze Age Selection Boom Rewrites Evolution
Step-by-step setup for the latest Perplexity Web MCP & CLI (v0.10.7) in OpenCode, using its free Big Pickle model as orchestrator to run Perplexity Pro "council" debates with models like GPT-4o, Gemini, and Claude—demoed on a humans-vs-AI prompt.
Free Perplexity LLM Council via OpenCode + MCP
Casual live stream of entrepreneur Chris Koerner fielding chat questions on starting AI agencies, IT consulting, real estate deals, lead-gen websites, and oddball ideas like a swirly-bottom muffin tin—while driving to Fort Worth, battling stream freezes.
AI Agencies & $100 Hustles: Chris Koerner's Live Advice
Step-by-step CLI tutorial for Vercel DeepSec, a security scanner that batches AI agent reviews (via Claude Code subscription) on large codebases to flag vulnerabilities. Demos setup commands (init/scan/process/report), tests on OWASP apps, and packaging as a Claude Code skill; resources at ailabspro.io.
DeepSec Structures AI Security Scans on Large Codebases
Tutorial on Vercel DeepSec CLI workflow with Claude Code: deepsec init (info.md prompt), scan (regex filter), process (batched agent review), report (JSON/MD), optional revalidation/export for codebase security scans. Found 3/10 OWASP issues in tests, misses runtime vulns.
DeepSec CLI Secures AI-Generated Codebases
Speaker argues Mozilla's use of Anthropic's Mythos—which found 271 vulnerabilities in Firefox—erodes trust in human-written code as the security baseline, urging a "golden refactor window" to prioritize code comprehensibility. Full writeup here.
Mythos' 271 Firefox Vulns Flip Human Code Trust
Nate Jones's monologue on Mozilla's experiment using Anthropic's Mythos to uncover 271 Firefox vulnerabilities, arguing it undermines trust in human-written code as the security baseline and shifts engineering focus to specs, comprehensibility, and AI-scale adversarial review. Full article w/ prompts.
Mythos Exposes 271 Firefox Vulns, Eroding Human Code Trust
Explains routing tasks between OpenClaw on Opus for high-stakes planning/review and Hermes on cheap models for routine execution, with a live demo building an HTML sponsor deal tracker via plan-execute-review. Builds to shared workspaces and task triage habits for cost savings.
Pair OpenClaw + Hermes to Halve AI Costs
Primeagen reacts to Bun's tweet announcing a Zig fork with 4x faster debug builds, blocked from upstreaming by Zig's strict no-LLM policy, then yaps on dev velocity, enforcement challenges, and AI's role in OSS futures.
Zig Rejects Bun's Fork Over LLM Policy and Flawed Speed Hack
Demo of Talkie, a 13B model trained solely on pre-1931 texts (newspapers, patents, books) to test AI reasoning vs. memorization without modern web contamination — shows predictions, slang, basic Python learning, and training hurdles like OCR and temporal leakage.
Talkie: 13B LLM with Pre-1931 Knowledge Only
Demo of NEO, a VS Code extension that plans/executes ML pipelines (synthetic data gen → model training → API/UI deployment) locally on tasks like chat moderation, with optional S3/HF/W&B integrations.
NEO Builds Full ML Pipelines in VS Code from One Prompt
Demo of NEO, a VS Code extension for automating ML workflows locally: takes a prompt to build a chat moderation model by generating synthetic data, training a baseline classifier, deploying an inference API, and creating a basic web UI, with setup and integrations explained.
NEO Automates Full ML Pipelines in VS Code from One Prompt
Sponsored tutorial for Freebuff, an ad-supported CLI AI coding agent using free models like GLM 5.1, DeepSeek v4 Pro, Kimi K2.6, and MiniMax M2.7. Covers npm install, model selection, subagents, and a live project demo with comparisons to Claude Code's speed and limits.
Freebuff: Free AI Coder 3x Faster Than Claude Code
Walkthrough of building a basic AI chat widget for a local business site: curate content (site, YouTube, Instagram) in Poppy, scrape products into a PDF with Claude Code, connect via Poppy API for queries, deploy to Vercel preview, and pitch as a $1k+ service with recurring upsells.
Sell Custom AI Agents to Local Biz: Claude + Poppy Stack
Creator's tiered ranking of their personal AI tools—from S-tier daily drivers (Claude Code, VS Code, Glido) and A-tier weekly ones (Codex, Claude chat, Hermes Agent, Perplexity, Grok) down to experiments and graduated tools—plus a decision framework for ignoring hype.
Copy This Lean AI Stack + Frameworks to Beat Overwhelm
Creator's subjective tier list of ~30 AI tools: S-tier daily drivers (Claude Code in VS Code, Glido), A-tier weekly (Codex, Perplexity, Grok), down to C-tier experiments and "graduated" ones he's dropped like ChatGPT and Cursor, plus a framework for sticking to a lean stack.
Nate Herk's Lean AI Tool Tier List
Advocates running Claude Code's terminal inside the Codex desktop app for a dual-tool workflow, with a walkthrough of Codex's UI/settings/plugins and pitches for the creator's Skool courses.
Run Claude Code in Codex for Dual AI Coding Power
Walkthrough of the Codex desktop app (OpenAI's coding tool), covering setup, UI, pricing, settings, plugins/skills, and a demo running Claude Code in its built-in terminal to use both together; includes pitches for the creator's paid masterclass and free community.
Use Claude Code + Codex Together for Best AI Coding
SaaStr CEO Jason Lemkin recounts deploying 20+ off-the-shelf AI agents (like Deli for a content-trained chatbot, Qualified for inbound, Artisan for outbound SDRs) by spending weeks training them daily—now handling support, sales Q&A, and bookings better than their small team. Key takeaway: constant iteration beats plug-and-play.
SaaStr's 20+ AI Agents: Train Hard, Replace Mediocre Humans
Quick demo of installing OpenAI's Codex Chrome extension on Brave, then using it to navigate the creator's LLM Council site (a Karpathy fork) and post the "hot dog sandwich" question for a model debate.
Codex Chrome Extension Automates Browsers via Natural Language
Podcast recaps Anthropic's DevDay agent features—Dreaming for scheduled memory review and pattern extraction, Outcomes for rubric-based output grading (+8-10% quality gains), and multi-agent orchestration— overshadowed by a brief SpaceX partnership mention.
Anthropic's Agent Harness Leap and SpaceX Compute Lifeline
News recap of DeepSeek-TUI, a Rust terminal agent powered by DeepSeek V4 that trended on GitHub with 10k+ stars. Covers the music/law student creator's story, viral buzz from devs, and features like sub-agents, context compression, and approval modes—no hands-on demo.
DeepSeek-TUI: Viral Open-Source Claude Code Rival
Casual on-the-ground chat from Anthropic's developer conference: two hosts recap announcements like the xAI compute deal, doubled usage limits, and new Claude Managed Agents features, speculate on OpenAI rivalry, and review the salmon bowl.
Anthropic's Compute Deal and Agents Challenge OpenAI
Live demo of "Marketing Brain," an Obsidian vault template driven by Claude Code prompts and DataForSEO API to pull competitor keywords (e.g., 18k uniques from top 10 sites) and generate 30/60/90-day SEO plans via the presenter's FLOW framework. Runs it on two client sites with setup walkthrough.
Marketing Brain: AI Vault for 18k Keyword SEO Strategies
Explains DESIGN.md, Google's markdown format for defining design systems (colors, typography, rules) that AI tools like Cursor/v0 can read to generate more consistent UIs from the same prompt. Includes a live demo, comparisons to Figma/JSON, and balanced pros/cons.
DESIGN.md Makes AI UIs Consistent and On-Brand