I use LLMs to learn complex topics

Aug 10, 2026 02:16 AM - 2 hours ago 1

Many engineers I cognize usage generative AI for galore functions, for illustration building PoCs, soul devices aliases dashboards, aliases moreover learning caller stuff. I personally find the style utilized by LLMs to explicate things difficult to follow. It's conscionable excessively simplistic and depending connected the number of emojis used, a spot annoying too.

How chips are made - explained successful a RollerCoasterTycoon-like simulation

While I was analyzing caller AI bottlenecks that mightiness slow down information halfway buildup, I realized location are galore aspects of spot accumulation that I do not know. Surfing the web, I asked myself what if location would beryllium a crippled to get you done the process of building a spot astatine a fab? For judge learning this measurement will stick, since you tin representation concepts pinch objects wrong the game. This is erstwhile I decided to effort it, and it really turned retired really well.

The flow

Instead of conscionable asking AI to explicate a topic, I usage the pursuing flow:

  • In scheme mode (using CC, aliases OpenCode) I inquire a exemplary to build the foundational knowledge for X topic.
  • I inquire it to reappraisal the accuracy of the knowledge guidelines it built successful the erstwhile step.
  • I proceed asking it to build a simulation of that taxable successful a low-poly, Rollercoaster Tycoon-like animation. I adhd immoderate UX elements arsenic well, for illustration the page needs to beryllium visible connected some ample and mini screens, person controls to extremity the travel whenever I want etc.
  • I past push it to a caller repo and alteration GitHub Pages for it.

The result

What you get is simply a beautiful animation that is 100% meticulous and free of hallucinations. For me, this method useful a batch amended than conscionable reference endless materials that I find connected Google, aliases trying to digest a bulleted database that is spat by a connection model.

I've done this specifically for learning spot building and motorboat it nether this website: ChipTycoon. You get to travel a cart from the infinitesimal erstwhile soil is collected, to the infinitesimal erstwhile a spot is finalized and delivered to a information center.

Visually, you tin travel the cart and spot really it changes too. Since it's low-poly, the specifications mightiness beryllium missing, but it's still a bully parameter for showing really the merchandise changes erstwhile it goes done the galore steps required successful the manufacturing process.

How to amended it further

Let's opportunity that the low-poly creation requires to overmuch immagination to really visualize what happened to the quartz soil heap aft it near the furnace. To toggle shape this into a much realistic representation, you tin usage my skill for transforming pictures into 3d objects, and representation the resulting objects to your simulation. This measurement you get much meticulous design.

Also, you tin adhd challenges to your simulation too. Trying to reply questions astir a erstwhile measurement successful the spot manufacturing process will thief you clasp the knowledge tremendously. Add intuitive puzzles excessively that will thief you study moreover better.

Check retired what different pages I created:

  • How rocket engines are made
  • How LLMs work
  • How F1 engines are built
  • How an EUV instrumentality is built
More