../
2026-02-17 Meeting
What I did?
- Refactor many parts of
pyverilogandply- Some parts are not well written
- They are still inheriting from
Object - No support of
systemverilog. May need one in the future if the assertions are syntactically wrong- But rn they seem fine
- Claude Code results on
apbi2c
| Module | Proven | Total | Cover % |
|---|---|---|---|
fifo |
27 | 27 | 100 |
apb |
22 | 22 | 100 |
module_i2c |
31 | 32 | 93% |
i2c |
19 | 19 | 94% |
Quirks
- Ran many commands by itself
ls, grep, head
- It didn’t know that
.tclfiles don’t support env expansion fifo.verrored out because it had a combinatorial loop, so I had to create a latch in-between- Each file took approximately 5 minutes
- Has no idea about jasper gold
- Created
loop_cutcommand out of thin air
- Created
- It automatically compacted my context window after the third file
- This was all done without a
CLAUDE.mdfile. Idk how much it will help but I just read paper concluding that it actually negatively affects performance across the standard LLM SWE benchmarks
43.k Tokens