Allen Hurff (@allenhurff) header
Allen Hurff (@allenhurff) profile picture

Digital Craftsman. Built platforms, teams, and ecosystems that actually scaled. Scar tissue earned. Still building what survives reality. Jaded optimist.

Baltimore, MD
Joined May 2007
Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)Photo by Allen Hurff (@allenhurff)
Data Centers versus Pitchforks Chart – at Baltimore, MD
1
2
147
Allen Hurff retweeted
I've got a pretty clear picture now of where we're headed next year, when Fable-class models become ubiquitous and cheap, and every enterprise is flooded with hundreds of new Fable-class AI employees. Spoiler: They create a constitutional legal system. yegge.ai/essays/fences-not-s…
51
46
41
611
98,713
Wake me up when I get to play. Till then, just shut up. – at Baltimore, MD
1
6
78
Are you like me and you have no interest in hearing a whole bunch of other people talk about "how great" AI models, ARE, that you don't have access to? It just doesn't land the same. In fact, it's offensive. It went from community to other people just bragging. And I know they're not even trying to brag. It just comes off that way now. – at Baltimore, MD
1
2
118
Allen Hurff retweeted
Can’t put the genie back in the GPU.
145
394
13
6,476
562,427
"1st" game that captured my HEART ❤️ as a kid was Sierra's `THE BLACK CAULDRON`. Always wanted to mess with the original source AND Space Quest ][ AND Police Quest. While it has been possible for a decade or so, not 100% in the original Ken and Roberta Williams source/toolset ...
1
62
Sure, why not 🤓
1
32
Oh, Wow! Already on to editing room one in the game!
27
1/3 Running the Apple AI / MLX / LLM Stack (local on macOS 27) youtube.com/watch?v=wykPErJ8…
1
78
2/3 So in NVIDIA world: NVIDIA GPU → CUDA → PyTorch / TensorRT-LLM / vLLM → model → app Apple "Local AI" (aka MacOS) world is roughly: Apple Silicon → Metal + unified memory → MLX → mlx-lm → model → app/agent
1
74
3/3 MLX Server Running / Responding (Local AI) mlx_lm.server \ --model mlx-community/Qwen3.5-4B-MLX-8bit \ --port 8080 \ --chat-template-args '{"enable_thinking":false}' curl -X POST http://127.0.0.1:8080/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model":"default_model","messages":[{"role":"user","content":"Hello!"}]}' also: http://127.0.0.1:8080/health http://127.0.0.1:8080/v1/models
65
Allen Hurff retweeted
Best YouTube Channels to Learn Languages & Communication: 1. English – English Addict with Mr Steve 2. Spanish – Dreaming Spanish 3. Japanese – Japanese Pod 101 4. Public Speaking – Charisma on Command 5. Writing – Brandon McNulty 6. Storytelling – Matthew Dicks 7. Debate Skills – Intelligence Squared 8. Body Language – Observe 9. Business Communication – Jeff Su 10. Vocabulary – Merriam-Webster 11. French – Français Authentique 12. German – Easy German 13. Mandarin Chinese – Yoyo Chinese 14. Arabic – Arabic Pod 101 15. Korean – Talk To Me In Korean 16. Italian – Italiano Automatico 17. Portuguese – Practice Portuguese 18. Hindi – Hindi Pod 101 19. Russian – Russian With Max 20. Sign Language – Bill Vicars 21. Accent Reduction – Rachel's English 22. Pronunciation – Sounds American 23. Grammar – English with Lucy 24. Listening Skills – TED Talks 25. Persuasion & Influence – Rhetoric & Persuasion 26. Presentation Skills – Toastmasters International 27. Confidence Building – Brian Tracy 28. Emotional Intelligence – Daniel Goleman Talks 29. Cross-Cultural Communication – Erin Meyer Insights 30. Speed Reading – Howard Berg 31. Memory & Vocabulary Retention – Loci Method 32. Slang & Informal English – English with Greg 33. IELTS Preparation – E2 IELTS 34. TOEFL Preparation – TST Prep 35. Debate & Critical Thinking – Jordan Peterson Lectures 36. Journalism & Reporting – Reuters Training 37. Screenwriting – StudioBinder 38. Poetry & Spoken Word – Button Poetry 39. Negotiation Skills – The Black Swan Group 40. Interviewing & Communication – Big Think
21
1,515
13
5,308
207,415
A doctor from 1985 seeing 2026 medicine would be stunned. mRNA vaccines. Cancer immunotherapy. CAR-T cells. Gene editing. AI protein design. Lab grown organoids. Personal genome sequencing. Robotic surgery. Blood tests that can detect cancer signals. Drugs that make people lose 20% of their body weight. We will make even bigger jump, just in the next few years.
35
89
9
556
33,511
OpenClaw: Main Markdown Self-Reflection (as monthly maintenance): This is something I run about once a month. ``` Review all of your core Markdown files in ~/.openclaw/workspace: * AGENTS.md * TOOLS.md * IDENTITY.md * USER.md * HEARTBEAT.md * MEMORY.md * SOUL.md Perform a full self-review against the current best practices in OpenClaw 2026.6.1 (and any newer patterns you know about). For each file: 1. Show the current content summary. 2. Identify what has improved in the latest version (structure, clarity, memory handling, dreaming, Codex/ACPX integration, agent presence, etc.). 3. Make specific, targeted recommendations for refinement. 4. Present any proposed changes as clear diffs or exact suggested sections. 5. Do NOT revert to generic defaults or "the mean". Preserve my unique presence, tone, fidelity, personality, and custom instructions at all costs. Only suggest improvements that enhance clarity, effectiveness, or alignment with the latest OpenClaw capabilities while keeping my level of depth and character intact. After reviewing all files, give me a final summary with: * What changed and why * The exact proposed updates for my approval * Any files that need no changes Only make changes if I explicitly approve them. Do not apply anything automatically. Begin the review now. ```
2
51
1/7 How I (repeatedly) update OpenClaw. Same Steps. Expected Outcome. After a (lite) backup prompt, I run this exact 5-step maintenance routine every single time. Keeps everything current, clean, and reliable. Here’s the full process:
1
8
32,768
6/7 5. Final Verification echo "== final verification ==" openclaw gateway status --deep --require-rpc --timeout 60000 openclaw status --deep --require-rpc --timeout 60000 openclaw plugins list | grep -i 'lossless' openclaw tasks audit --json Only stop when everything comes back green.
1
63
7/7 This whole routine takes me under 5 minutes and has saved me from countless silent failures. Same steps. Expected outcome. Every time. Still building what survives reality.
20