
MenuThe loop in one line:
Claude scopes and specifies → Local LLM writes each module → Claude reviews each module → Local LLM fixes → Claude does integration review → Warp builds, flashes, and...
And in plain English, which will use more tokens , but is way easier to understand
Same example: WiFi-connected ESP32 that monitors a digital input, logs state changes with a timestamp to internal...
So to better explain, a worked example of using the three systems:
Stage 1 — Claude: Scope and Spec
System prompt (cached):
ROLE: technical architectOUT: specification document +...
you asked why two ai systems would "talk to one another".
So thats effectively asking why I use more than one AI system.
here is the simple answer - in reality its more complex and sometimes...
re Json Tags:
Is there a standard?
Not really, truth is yoiu can make your own and your ai will adapt to it. DO use plain english, its what the system was trained on
Anthropic's own prompt...
Because FREQOUT is using PWM with the ability to output two tones at the same time, so filtering is required to produce a, relatively clean, sinewave-looking tone.
On the otjher hand, SOUND...
Hello :)
AI has no answer.
even at 64mhz, FREQUOUT provides very low volume, distorted audio, while SOUND, even on 4Mhz 628A chip, produces loud and clear sound.
This is not chip or hardware...
Thank you Lester for the full guidlines on using AI in a better and efficient way.
I was very temted many times to reply politely but got self constrainted as I do understand that it is a machine...
Garbage In, Garbage OutWe all know the rule. Garbage in equals garbage out. It has not changed. AI does not break that rule. AI enforces it.
Do not blame the model when the code comes back wrong....

Re: Has anyone tried AI with PICBASIC
Thank you Lester for the details and your efforts to explain.
Ioannis Yesterday, 12:39Makes sense why to use many models.
Great tools we have now. Have to make clever use of them as you showed with detailed examples....