Kurka Labs · Project COHO · Seminar 1
Introduction to AI
A practical orientation for curious professionals and entrepreneurs
NameDate
This hour is a map—not a mastery course. The goal is to distinguish the parts of modern AI, see what the tools can and cannot do, and leave with a safer way to explore.
By the end:
You should be able to compare tools, give an AI better context, and recognize when human judgment must remain in charge.
You should be able to compare tools, give an AI better context, and recognize when human judgment must remain in charge.
The one-hour route
0–5
Welcome & the map
What have you tried? What worked, failed, or concerned you?
5–13
AI in plain language
Pattern-driven generation; generation is not the same as retrieval.
13–22
Vendors, products & models
Why “the AI” hides several different layers and choices.
22–32
What chatbots can & cannot do
Broad transformation abilities without guaranteed truth or judgment.
32–42
Browser versus terminal
A conversation interface compared with an agent inside an environment.
42–49
Cloud versus local
Capability, convenience, privacy, control, cost, and maintenance.
49–56
Three demonstrations
Two chatbots, a real file, and the supervised PubX workflow.
56–60
What comes next
From chatbot to collaborator, followed by questions and feedback.
The five-part map
VendorWho provides it?
ProductWhat service?
ModelWhat engine?
InterfaceHow do you use it?
LocationWhere does it run?
A fluent answer is a generated result—not proof. For important claims, ask what source can be inspected and how the result can be checked.
1 / 2
Working guide
Use the tool. Keep the judgment.
Most useful idea
Browser ↔ terminal
| Browser chatbot | Terminal agent | |
|---|---|---|
| Best at | Conversation, uploads, visual interaction | Files, commands, tests, connected artifacts |
| Sees | What you provide or connect | What its permissions allow |
| Risk | Unverified or generic output | Changes made in the wrong place or scope |
Cloud ↔ local
| Cloud | Local | |
|---|---|---|
| Advantage | Power, convenience, rapid improvement | Control, privacy, direct environment access |
| Tradeoff | Provider dependence and data boundaries | Hardware, setup, upkeep, smaller models |
| Reality | Many useful systems will combine both. | |
Objective + Context + Constraints + Review = a better request
A simple trust ladder
Explore freelyBrainstorm, outline, transform, compare, and ask questions.
Verify before relyingFacts, citations, calculations, legal or medical claims, and current events.
Keep human authorityPublishing, spending, deleting, sending, signing, and consequential decisions.
Try this: I am [role/context]. I need [objective] for [audience]. Preserve [constraints]. Ask before [consequential actions]. Give me [format], identify uncertainty, and tell me how to verify the important claims.
Notes, questions, or an experiment to try
Before you leave
- What became clearer?
- What still feels confusing or unbelievable?
- Which demonstration should we explore next?