Advice for anyone (re)learning how to build software with AI tools:
Stop buying someone else's "perfect AI skills repo." There's no universal senior engineer prompt.
Map your own workflow, isolate what's actually automatable, and build narrow skills that fit.
Deterministic logic + LLM magic only works when YOU design it.
This vial contains a new drug called PAC-832, which I recently invented to treat Alzheimer’s disease. It is the world’s first selective GalR1 antagonist.
I designed and synthesized PAC-832 in a chemistry lab I built in my garage. (1/16)
DeepSeek just released DSpark for V4 Flash & Pro, a new speculative decoding method boosting throughput by 51% to 400%!
DS also showed DSpark works well for other models like Gemma & Qwen
Github: https://t.co/EGVYpc1kcK
Paper: https://t.co/TaBMRVlaW9
HF: https://t.co/289jVU2pxh
NEW: Airbound is building 5-minute drone delivery across India.
The Bangalore startup is building ultra-light autonomous drones to move medicine, lab samples, and small packages faster than scooters in traffic.
• Raised $8.65M led by @lachygroom
• Drone weighs ~5.5 pounds and can carry ~2 pound payloads
• Flies up to ~25 miles at ~40 mph
• Completed 700+ medical delivery flights between clinics and labs
• Planning up to 10,000 drone flights per day in Andhra Pradesh
@Airbound_Aero is targeting delivery costs as low as $0.01 at scale.
A curated list of LLM papers, frameworks, and tools, tracking the field from transformers to modern reasoning models.
- 300+ milestone papers from GPT-1 to DeepSeek-R1
- Training frameworks like Megatron-LM and inference tools
- Open LLM checkpoints, leaderboards, and evaluation resources
- Tutorials, courses, and books for learning LLM development
Introducing Mixture of Agents 2.0 in Hermes Agent.
Combine any provider's models into a mixture of your own. Access your presets as if it were a normal model in Hermes.
Big improvement in our soon-to-release HermesBench against opus and gpt-5.5 with MoA using Opus & GPT together.
16 parallel runs of Gemma 4 26B A4B on a single NVIDIA DGX Spark!
Pushing 18 tok/s per instance and a 300 tok/s aggregate. It can even hit 32 parallel runs.
This level of concurrency highlights how efficient the architecture is.
My X feed this week was wild. Here's what stood out:
Open source is eating everything. Activepieces replacing Zapier. Presidio catching PII before it leaks into your LLM pipelines.
AI agents are getting serious. Loop Library launched. Hermes shipped batch memory. The harness over model debate is heating up.
ComfyUI is changing video generation. Qwen-4.0-max and GLM-5.2-Turbo incoming.
Also: humanoid robots market map, a developer replacing Electron bloatware, and a Linux VM running in your browser.
What a week.
#AI #OpenSource #AIAgents #DevTools #BuildInPublic
Everyone is debating which AI model is best.
Wrong question.
The model doesn't matter if your workflow is broken. The teams winning with AI right now aren't using the smartest model. They're using the best harness.
Prompting is a skill. But system design is the superpower.
#AI #AITools #TechLeadership #Productivity #BuildInPublic
Most people think AI is magic. It's not.
It's pattern recognition at scale.
The real skill isn't prompting — it's knowing what problems are actually worth solving.
Stop chasing the latest model. Start understanding the problem you're trying to fix.
That's where the value lives.
#AI #MachineLearning #ArtificialIntelligence #TechLeadership #BuildInPublic
Matt Pocock just explained why everyone is obsessing over the wrong thing
it's not the model, it's the harness
watch this before everyone else figures it out