🤖 #unautomatable | Owner/CEO @AIMakerspace
Columbus, Ohio
Joined April 2010
- Tweets1.4K
- Following44
- Followers416
- Likes1.1K
Greg Loughnane retweeted
Yann LeCun says LLMs are strongest in domains where language itself is the substrate of reasoning, like math and code
They can solve problems, prove theorems, and write programs — but they are not creative mathematicians, software architects, or computer scientists
"their role is to help humans build"
Greg Loughnane retweeted
.@PalmerLuckey’s advice for young founders:
“If you want to work on technology, work for somebody else. If you want to work on nonstop bullshit—start a company.”
“It became clear there were people much better than me at certain things.”
“I had to stop doing the things I liked, and start doing things I was uniquely able to do— running the company, doing fundraising.”
"Marketing is a revenue function, not a creative one."
A great friendly reminder from @ryandeiss
The best judge for your LLM application is your user.
The LLM that judges your LLM application must have its judgement judged (even if only downstream) by a human user.
Abstraction does not remove the need for contact with reality and survival over time.
#unautomatable
(Agile?) Context Engineering for Agents & Multi-Agent Systems, with @llm_wizard, by @AIMakerspace & @MavenHQ
youtu.be/LuA4DLbZIpI
Big Takeaways 👇
Foundational Context Engineering references with big main ideas behind each👇
- (March 30, 2025) 12-Factor Agents
Repo: github.com/humanlayer/12-fac…
Talk: youtube.com/watch?v=8kMaTybv…
💡Big idea: own your context window and treat it like prime real estate; minimize “context utilization”
- (February 2001) Agile (Context) Manifesto
Tight agent handoffs + HITL over sloppy autonomy
Working application outputs over verbose agent conversations
Goal-aligned agent teams over rigid templates
Frequent compaction & re-planning
Orig: agilemanifesto.org/