i invented a new word: "slopgrace". it combines "slop" and "disgrace". i

I've never even really considered using AI for writing kernel patches.

We could go through the usual reasons, like simply congitive abilities to be a gatekeeper for anything but there are also cost-benefit measures that simply make using AI a bad investment in this particular area of expertise.

Writing the code takes about 1-5% of the all time spent on a kernel patch. It is still really important meditation time too, as then you can go through your plan/strategy slowly and make sure every brick is in the right place. I notice tons issues while writing code even if the overall plan is solid as **** (... and this is why BTW spec-driven agent loop development is complete and uttermost nonsense).

By cutting that 1-5% there is at least no financial at all, and the drawback is a constant ffux of kernel bugs because of *mistakes*, not unexpected conditions. Even if it was 0,1% that a frontier LLM generates a kernel bug, it is 0,1% too much. And even if I can human-check and test it (which I would, it is not at all the same than the quality time spent browsing the kernel tree :-)

Sent what is possibly (even by my standards) the longest mail I've even sent on-list.

But it addresses really serious concerns about the AI trend, as well as the series reviewed in particular.

And I think it really matters.

https://lore.kernel.org/all/aj9yrlB0TrlYCLlf@lucifer/

libttk, a TUI toolkit written in C has been part of various experiments over the last few years. Never got released because I haven't had really any real application for it. CCD is the most recent of them. I think CCD will get released some day once I have time to polish the rough edges :-)

I try to use Rust crates that use common C libraries as much as possible because Rust crates tend to become non-upgradable fast if you want to retain some sane MSRV.

E.g., right now I cap to version 1.85 as it is what Debian Trixie has. Usually those crates that bind to C libraries are more conservative also with MSRV.

victory, now that stupid logo image is also showing up ;-)

Made a splash with Typst.

Ugh, had forgotten gnome-keyring from Puu OS as I was fighting with homectl and fscrypt v2.

And it requires GCR 3!?

GCR goes ATM in 4.x series.

I already see dreams getting my fresh NAS drives :-) Theoretically getting em some time in October. They used to be commodity gear. Now they are like the ***ing Playstation or something.

gbrain - the postgresql of agents ;-)

garry tan did it again

New subcommand in Goosedump called ‘delete’:

❯ goosedump delete 'gemini*'

gemini:session-2026-03-01T07-01-be63c197
gemini:session-2026-03-01T06-03-3289e66b
gemini:session-2026-03-01T04-08-b988a3ae
gemini:session-2026-02-28T14-42-821d5fcd
gemini:session-2026-02-28T14-14-7205125e
gemini:session-2026-02-28T12-11-f3ecf0c8
gemini:session-2026-02-28T06-44-b2a2bee6
gemini:session-2026-02-26T20-57-e195e041
gemini:session-2026-02-25T20-08-5467631c
gemini:session-2026-02-24T10-10-d1bcd2c4
gemini:session-2026-02-22T17-41-55c58931
gemini:session-2026-02-22T08-10-a78d04be
gemini:session-2026-02-22T07-56-a78d04be
gemini:session-2026-02-16T23-13-0a78af81
gemini:session-2026-02-12T14-09-dc5f9bc5
gemini:session-2026-02-12T14-07-dfdcbed1

"goosedump delete '*' deletes all sessions from every installed coding agent giving a convenient way to wipe a laptop :-) It does have ‘–dry-run’ too.

I’ll implement later on also ‘import <provider> <glob>’, and then import + delete will effectively give move semantics.

IMHO, the name really nails this tool and its purpose.

I think I should do some kind of output filtering to make opencode-landstrip output just contain the tool command :-)

{ exec 3<>/dev/tcp/127.0.0.1/45811 ; } 2>/dev/null && '/home/jarkko/.cache/opencode/packages/opencode-landstrip@latest/node_modules/@landstrip/landstrip-linux-x64/bin/landstrip' '--trap-fd' '3' '-p' '/tmp/opencode-landstrip-df17xp/policy.json' '/bin/zsh' '-lc' 'env TARGET_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/target" BINARIES_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/images" HOST_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/host" BR2_EXTERNAL_PUU_PATH="/home/jarkko/work/codeberg.org/puu/puu" PUU_VERSION="2026.05" SOURCE_DATE_EPOCH="1735689600" "/home/jarkko/work/codeberg.org/puu/puu/board/puu/amd64/post-image.sh"' || bash -c 'exec 3<>/dev/tcp/127.0.0.1/45811 2>/dev/null && exec "$@"' bash '/home/jarkko/.cache/opencode/packages/opencode-landstrip@latest/node_modules/@landstrip/landstrip-linux-x64/bin/landstrip' '--trap-fd' '3' '-p' '/tmp/opencode-landstrip-df17xp/policy.json' '/bin/zsh' '-lc' 'env TARGET_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/target" BINARIES_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/images" HOST_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/host" BR2_EXTERNAL_PUU_PATH="/home/jarkko/work/codeberg.org/puu/puu" PUU_VERSION="2026.05" SOURCE_DATE_EPOCH="1735689600" "/home/jarkko/work/codeberg.org/puu/puu/board/puu/amd64/post-image.sh"' 2>/dev/null || '/home/jarkko/.cache/opencode/packages/opencode-landstrip@latest/node_modules/@landstrip/landstrip-linux-x64/bin/landstrip' '-p' '/tmp/opencode-landstrip-df17xp/policy.json' '/bin/zsh' '-lc' 'env TARGET_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/target" BINARIES_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/images" HOST_DIR="/home/jarkko/work/codeberg.org/puu/puu/build/puu_amd64/host" BR2_EXTERNAL_PUU_PATH="/home/jarkko/work/codeberg.org/puu/puu" PUU_VERSION="2026.05" SOURCE_DATE_EPOCH="1735689600" "/home/jarkko/work/codeberg.org/puu/puu/board/puu/amd64/post-image.sh"'

https://www.reuters.com/world/europe/russian-held-sevastopol-without-power-after-ukraine-strikes-moscow-installed-2026-06-24/

❯ tpm2sh algorithm
ecc-bn-p256:sha256
ecc-nist-p256:sha256
keyedhash-hmac:sha256
keyedhash-null:sha256
keyedhash-xor:sha256
rsa-2048:sha256

tpm2sh 1.0

https://crates.io/crates/tpm2sh

Meson's wraps are awesome :-) I've learned about this feature only recently. Goodbye submodules...

Something I just learned: if you ask filesystem permission do it BREADTH-FIRST, NOT depth first ;-) It takes having a ~/.cargo/registry hierarchy to realize this.

Landstrip is like browser tab type sandbox for command-line.

I've actually wanted something like that to restrict e.g. complex potentially destructive scripts and stuff like that long before AI psychosis but coding agents make such great test case for stress testing, as they stochasticaly make wrong choices all the time.

Why they might suck makes them great for this project.

I found DDR3's from my drawer? Anyone? Affordable prices ;-) Or maybe I should let Sam and Elon fight for these to get the best bid.

Wondering if I should get insurance for my RTX 4090

Newest victim of the AI hype: 20-year old DDR2 memory

https://www.tomshardware.com/pc-components/dram/ddr2-memory-prices-jump-up-to-60-percent

Great all PRs went in w/o complains. Last round I had some issues, so had a little bit of extra stress :-)

I have now two next branches:

1. for-next-keys
2. for-next-tpm

I always send (1) first as it has larger spread (every system that boots Linux has keyring) and after that (2). Before I had them mixed up, which was a really bad idea (or there was no thinking at all tbh), so had to reorg...