So this is rough plan, i am not sure whether i will be in bangalore or in ahemdabad or where at that time, but idea is to complete them as much as possible, cause all things i want them to happen!
Every project needs to be somehow contributing to community or other new persons to learn more, fun, and something which i learn a lot and enjoy with good twist, and also idea is to deploy and cary out as many experiments as we can at scale so we can have more information about things, we never know what may work or what not, and what good things can come out of any experiment.
maximising surface area of luck and doing everything in order to achieve it.
projects:
- bytebeat
- NES Emulator
- hypervisor and OS
- JSON parser
- Toy Autodiff Engine
- raytracing
-
fastbrew - open source version of https://yutori.com/ (https://arxiv.org/html/2411.00820v1)
- open source version of https://generalagents.com/ace/
- voxel diffusion
- fully automated d2c store
- recommendation engine with simulator
- mech inter whisper
- llm inference engine
- Physically Based Rendering:From Theory To Implementation
- open thesis(open source version of anti thesis)
- SOTA computer use model
- open source version exa
- open source turbo puff
- shadowfax
- Supervised Policy Learning for Real Robots
- fully automated dark factory for manufacturing
- https://jajoo.sh/rtsci
- 31b1 kind of animation video for any research paper, but more than his video his explnation is good, need other kind of video style to make it good
- Incremental Sheet Formingindian version of https://machinalabs.ai/
- indian founders need this very badly: https://www.thefuturefactory.ai/
- Minecraft Robot Self driving skateboards trained on RL and CV models
Books:
- Nand2Tetris
- Jermy Kuan Maths book
- fast ai deep learning
- karpathy zero to hero
- https://fleuret.org/dlc/
- competitive programming book
- How to Scale Your Model
- reverse engineering for beginners
- reinforcement learning and simulators(they are good thing see: https://github.com/TorchSim/torch-sim)
- skiena algorithm design manual
- shenzhen-io
- SICP
- Ai is good by ERIC jang
Skills:
mastering them
- rust
- python
- pytorch
- jax
- assembly
- maths
- reverse engineering
- skateboarding
- video editing
- electric guitar
- cooking
- writing essays and thoughts more on substack(ability to express my thoughts more easily)
goals/metrics:
- 10k followers on linkedin and twitter
- 5k subscriber on newsletter
- 20 books finished reading
- 20 blogs written
- 20 high quality projects done
- making $1M dollars
- reaching out to 100 people
companies on the radar(grab opportunity if given or create):
- Nvidia recommendation engine team(https://developer.nvidia.com/blog/author/eoldridge/)
- world models at Meta(Yann LeCun)
- Anthropic(Mech Intr)
- Deepmind(Mech Intr)
Movies:
-
“Stories We Tell” (2012) directed by Sarah Polley:
- Context: A documentary that investigates the director’s own family secrets. It creates a “meta-narrative” about how different people remember the same event differently, directly paralleling the themes of truth and fabrication in F for Fake.
-
“Close-Up” (1990) directed by Abbas Kiarostami:
- Context: Based on a true story about a man who impersonated a famous film director. The movie features the actual people involved re-enacting the events. It is a masterpiece of deconstructing the “truth” of cinema.
-
“The Gleaners and I” (2000) directed by Agnès Varda:
- Context: Varda is the godmother of the cinematic essay. This film is a wandering, personal exploration of “gleaning” (collecting leftovers), but it is structured by her own curiosity rather than a traditional plot, much like a video essay.
-
“Exit Through the Gift Shop” (2010) directed by Banksy:
- Context: Startlingly similar to F for Fake. It starts as a documentary about street art but evolves into a question of who is conning who. It questions the value of art and the “expert” opinion.