with practice comes perfection (@willdoingthings) header
with practice comes perfection (@willdoingthings) profile picture

with practice comes perfection

@willdoingthings

navigating between the Scylla of overusing metaphors and the Kraken of mixing them up.

 Dream Zanarkand
Joined September 2012
  • Tweets23.2K
  • Following261
  • Followers710
904Media
Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)Photo by with practice comes perfection (@willdoingthings)
Opus can also add weights to KataGo that actually help it play better, of course:
47
clearly a neural network can interpret and modify or add to another neural network to give it new skills, such as drawing ravens in this case. Opus 5.5 doesn't have some native modality for model weights, but it does have solid, superhumanly fast facility with the standard tools.
1
87
here's a KataGo with a grafted-on "Raven bypass" that, whimsically, makes it draw a raven with its stone placements, then continue playing normally.
1
148
inspired, I asked Opus to improve, via grafting, an older KataGo that struggles with a type of position, and it did. this isn't so surprising in that Opus is using standard, known interpretability techniques (linear probes, logit lens), and the bolted-on behavior is algorithmic.
Rikiya Takehi (@rikiyatakehi) profile pictureRikiya Takehi@rikiyatakehiNov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Image from Rikiya Takehi (@rikiyatakehi)
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
2
1
233
Nathan dupes his subjects for sublime comedy but also his audiences for cheap drama. cf. The Rehearsal, season 2: Nathan doesn't sincerely care about aviation safety, yet he presents his (in)sincerity as a live question even metatextually. it's a meaningless, laughless schtick.
1
156

No more items