Tessera

@serpent7776.bsky.social

The Syntax Sorcerer aka I like programming languages.

Running your Claude Code with --dangerously-skip-permissions, but worried that it can rm -rf your precious cat pictures collection? I'm building a Claude Code sandboxer that limits potential damage radius to the project repo you're working on. github.com/serpent7776/...

GitHub - serpent7776/aye-buddy: Your AI buddy, a Claude Code sandbox

Your AI buddy, a Claude Code sandbox. Contribute to serpent7776/aye-buddy development by creating an account on GitHub.

github.com

@serpent7776.bsky.social · 7d ago

#TIL in rust you need to wrap field expression in parens to call it as a function.

$Note Wrap the field expression in a parenthesized expression to use it in a call expression. let holds_callable = HoldsCallable { callable: || () }; // Invalid: Parsed as calling the method "callable" // holds_callable.callable(); // Valid (holds_callable.callable)(); Examples: mystruct.myfield; foo().x; (Struct {a: 10, b: 20}).a; (mystruct.function_field)() // Call expression containing a field expression https://doc.rust-lang.org/reference/expressions/field-expr.html#r-expr.field.not-method-call$

@serpent7776.bsky.social · last wk.

The dangers of Postgres subtransactions

The subtransaction cache limit

A subtransaction is a transaction nested inside a top-level transaction. Applications create them explicitly with SAVEPOINT. PL/pgSQL also creates them for blocks that contain an EXCEPTION clause.

Each PostgreSQL backend keeps the assigned, non-aborted subtransaction IDs for its current top-level transaction in shared memory. This list is called the subtransaction cache and holds up to PGPROC_MAX_CACHED_SUBXIDS IDs, usually 64.

https://planetscale.com/blog/the-dangers-of-postgres-subtransactions

Explicit savepoints are not the only way to reach the limit. PL/pgSQL uses a subtransaction for every block that contains an EXCEPTION clause. The following loop creates more than PGPROC_MAX_CACHED_SUBXIDS subtransactions without issuing SAVEPOINT directly:

postgres=> ROLLBACK;
ROLLBACK

postgres=> DO $$
BEGIN
FOR i IN 1..70 LOOP
BEGIN
INSERT INTO subxid_test VALUES (i);
EXCEPTION WHEN OTHERS THEN
RAISE;
END;
END LOOP;
END
$$;
DO

https://planetscale.com/blog/the-dangers-of-postgres-subtransactions

Warning

A single transaction that accumulates more than PGPROC_MAX_CACHED_SUBXIDS assigned, non-aborted subtransaction IDs can slow queries across the entire PostgreSQL cluster, including sessions that do not use subtransactions themselves.

https://planetscale.com/blog/the-dangers-of-postgres-subtransactions

@serpent7776.bsky.social · 2w ago

dockerd: "Container failed to exit within 10s of signal 2 - using the force"

Darth Vader Uses Force on Ship

Alt: Darth Vader Uses Force on Ship

static.klipy.com

@serpent7776.bsky.social · 2w ago

This is supposed to be the AI frontier? ■ Failed to branch before the selected prompt: the selected prompt was not found in the persisted thread

@serpent7776.bsky.social · 2w ago

Codex is funny, it autonomously decided to ship code to prod on Friday night.

@serpent7776.bsky.social · 2w ago

Kimi K3 "escaping" sandbox

@serpent7776.bsky.social · 3w ago

Your local agent should have access to the same tools, but not the same environment as you.

Local agents inherit laptop access

Most local coding-agent setups begin with a reasonable assumption: the agent should have access to the same tools as the developer. In practice, though, a developer laptop usually contains more access than one task needs.

The most obvious risk is credential exposure. A laptop may have deploy keys, GitHub tokens, cloud profiles, database URLs, package manager tokens, SSH agent access, and one-off secrets that were never meant to be part of an agent task.

An agent does not need malicious intent to expose a secret. It can print an environment variable in a log or add a credential to generated code. It can also run a diagnostic command that shows sensitive data.

The next risk is network access. A laptop can usually reach internal tools, staging systems, admin panels, and private package registries. An agent troubleshooting a failing test may make HTTP requests, run a CLI, fetch a dependency, or follow a link without understanding which systems are sensitive. If agents can reach all the same places as the developer, "just let agents investigate" becomes a security concern.

https://www.superconductor.com/blog/remote-sandboxes-agent-security

@serpent7776.bsky.social · 3w ago

Simdvis now visualises VPMOVMSKB. serpent7776.github.io/simdvis/pmov...

Screenshot of an interactive visualization explaining the AVX2 VPMOVMSKB instruction. The interface shows a 256-bit YMM register split into 32 color-coded bytes (src), with each byte displayed in hexadecimal and binary. Green boxes below indicate the extracted most significant bit (sign bit) from each byte, forming a 32-bit mask in register r32. Arrows connect source bytes to their corresponding output bits. The example result is r32 = 0x17D4EA0B, with accompanying text stating that 16 of the 32 sign bits are set and TZCNT = 0, indicating the first byte with its sign bit set is src[0]. Control buttons for editing byte values (0x00, 0xFF, alt, rand, 00/FF) appear in the upper-right.

@serpent7776.bsky.social · 3w ago

Simdvis now visualises KSHIFT* and KUNPCK* instructions.

Screenshot of an interactive visualization for the AVX-512 KSHIFTLW instruction. The interface shows a 16-bit source mask (k2 = 0x91E3) in purple and the resulting destination mask (k1 = 0x8F18) in green after performing a left shift by an immediate value of 3. Curved arrows illustrate how each bit moves three positions toward the most significant end, with bits shifted past the 16-bit boundary discarded and zeros inserted into the least significant positions. Controls at the top allow changing the shift direction and immediate value, while the bottom summarizes the input and output mask values in hexadecimal and binary.

https://serpent7776.github.io/simdvis/kshift.html#kshiftlw

@serpent7776.bsky.social · 4w ago

Wait, rust atomics don't have .wait like c++ atomics do?

@serpent7776.bsky.social · 4w ago

some people like wasting their time.

Strict No LLM / No AI Policy

No LLM-generated content, whether it be code or prose.

No paraphrasing LLM-generated content.

No LLMs for editing, including fixing spelling or grammatical errors.

No LLMs for translation. English is encouraged, but not required. You are welcome to post in your native language and rely on others to have their own translation tools of choice to interpret your words.

No LLMs for brainstorming and then sharing the results of that brainstorming, even if you create the prose. If you use a chatbot to give you advice on a comment on the issue tracker, that comment is unwelcome.

No LLMs for finding bugs.

No talking about use of chatbot/LLM services.

https://ziglang.org/code-of-conduct/

@serpent7776.bsky.social · 4w ago

#zig: rejecting LLMs because of quality reasons also zig: HTTP client fails to make HTTPS requests via a proxy This was opened 2 years ago, how is this serious? github.com/ziglang/zig/... codeberg.org/ziglang/zig/...

HTTP client fails to make HTTPS requests via a proxy · Issue #19878 · ziglang/zig

Zig Version 0.12.0 Steps to Reproduce and Observed Behavior Run a local proxy server, say mitmdump Setup environment variables for http(s) proxy export https_proxy=http://127.0.0.1:8080 Make sure t...

github.com

@serpent7776.bsky.social · 4w ago

Simdvis now visualises *PCMP* instructions. serpent7776.github.io/simdvis/vpcm...

Screenshot of an interactive visualization for the AVX-512 instruction VPCMPB (VPCMPEQB). The interface compares two 128-bit vectors (src1 and src2) as signed bytes using the equality predicate (EQ). Sixteen byte values for each source are displayed in labeled boxes, with matching values highlighted in the resulting mask k1. The output mask shows bits set at positions 0, 6, 8, and 14, producing the hexadecimal result 0x4141 (binary 0b0100000101000001). Configuration controls across the top specify signed comparison, equality predicate, XMM (128-bit) width, and no mask.

@serpent7776.bsky.social · 4w ago

who needs docker when I have bubblewrap + landlock + systemd-run + pasta

@serpent7776.bsky.social · 4w ago

PR_SET_CHILD_SUBREAPER - set/unset the "child subreaper" attribute of the calling process

A subreaper fulfills the role of init(1) for its descendant processes.
When a process becomes orphaned (i.e., its immediate parent termi-
nates), then that process will be reparented to the nearest still liv-
ing ancestor subreaper. Subsequently, calls to getppid(2) in the or-
phaned process will now return the PID of the subreaper process, and
when the orphan terminates, it is the subreaper process that will re-
ceive a SIGCHLD signal and will be able to wait(2) on the process to
discover its termination status.

https://man.freebsd.org/cgi/man.cgi?query=PR_SET_CHILD_SUBREAPER&apropos=0&sektion=0&manpath=Debian+13.5.0

@serpent7776.bsky.social · 4w ago

I prepared a plan and now I'm doing 6th cycle of plan reviewing, and codex still finds high priority issues to fix. halp

@serpent7776.bsky.social · 4w ago

lstopo - Show the topology of the system

Diagram of a computer's CPU topology. A single machine with 23 GB of memory contains one NUMA node (23 GB), a shared 6 MB L3 cache, four 512 KB L2 caches, four 48 KB L1 data caches, and four 32 KB L1 instruction caches. Below the caches are four CPU cores (Core #0–#3). Each core has two processing units (hardware threads): Core 0 has PU #0 and #1, Core 1 has PU #2 and #3, Core 2 has PU #4 and #5, and Core 3 has PU #6 and #7.

@serpent7776.bsky.social · 4w ago

Rust v Fil-C is essentially dick length measuring contest where each tries to shout louder than the other while trying the hardest they can to ignore the arguments the other side has.

@serpent7776.bsky.social · last mo.

Rust and Zig: shits on each other Me: writes a project in Perl and don't care

@serpent7776.bsky.social · last mo.

Everybody is quoting Uncle Bob not reading the code any more, but no-one asking what kind of software has he built this way.

@serpent7776.bsky.social · last mo.

Standard #SQL can do row pattern matching now? Crazy. Too bad almost nobody supports that.

The example demonstrates a row pattern that matches one or more consecutive new messages (X), while allowing optional read markers at the beginning or end of a match and between groups of new messages.

SELECT id, marker
FROM msg
MATCH_RECOGNIZE (
ORDER BY id
ALL ROWS PER MATCH
PATTERN ( (^read)? new+ (read new+)* (read$)? )
DEFINE
new AS (marker = 'X')
) T
ORDER BY thread_id, id

https://modern-sql.com/feature/match_recognize

@serpent7776.bsky.social · last mo.

Those are exactly the same things that people benefit from, it's just that for years nobody cared.

Haskell should be unusually well positioned to be relevant to AI-enabled engineers and agents. Type safety can be a huge advantage for LLM-generated code if the compiler is helping the agent converge quickly. But that is not the same thing as optimizing the language for humans writing code by hand. Agents have different bottlenecks. They are cheap at generating code and expensive when blocked. They benefit from fast feedback, clear examples, low setup friction, and errors that help them repair the code quickly.

https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html

@serpent7776.bsky.social · last mo.

Nice #perl trick: /r regex modifier: perform non-destructive substitution and return the new value

$# mutates $x, returns COUNT removed chomp($x) # leaves $x alone, returns the trimmed COPY my $y = $x =~ s/\n\z//r;$

@serpent7776.bsky.social · last mo.

Here's my thoughts after reading Animal Farm by George Orwell. hive.blog/hive-180164/...

Animal Farm: A Fairy Story — Hive

Animal Farm by George Orwell by serpent7776

hive.blog

@serpent7776.bsky.social · last mo.

How insane is it to disable terminal mouse clicks if user disabled telemetry/non-essential traffic option in claude code? github.com/anthropics/c...

[BUG] DISABLE_TELEMETRY / DO_NOT_TRACK / CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC silently disable terminal mouse click interaction · Issue #73320 · anthropics/claude-code

Preflight Checklist I have searched existing issues and this hasn't been reported yet This is a single bug report (please file separate reports for different bugs) I am using the latest version of ...

github.com

@serpent7776.bsky.social · 2mo ago

simdvis now visualises VPMOVSX/VPMOVZX instructions serpent7776.github.io/simdvis/vpmo...

Screenshot of an interactive visualization for the AVX-512F VPMOVSXBD instruction in a dark-themed interface. The top section shows a 128-bit source vector containing 16 byte values (for example, 0x91, 0x32, 0x93, 0x34), with byte index [9] (0x3A) highlighted. The lower section shows the 512-bit destination vector, where each source byte has been sign-extended into a 32-bit integer. Bytes with the high bit set become values such as 0xFFFFFF91, while positive bytes become values such as 0x00000032. A connecting highlight traces the selected source byte 0x3A to destination element [9], which is 0x0000003A. A caption at the bottom explains: dst[9] = 0x0000003A ← sign-extend(src[9] = 0x3A); high 3 bytes = 0x000000 (sign bit = 0).

@serpent7776.bsky.social · 2mo ago

wait, I can write zig code that parses and manipulates zig code using std.zig?

@serpent7776.bsky.social · 2mo ago

"You're not building big enough"

Presentation slide from CascadiaJS 2026 showing keynote speaker Theo Browne at a podium on the left and a large black slide on the right with the handwritten message, “You're not building big enough.”

@serpent7776.bsky.social · 2mo ago

The only sane fsync behaviour.

FreeBSD’s Approach

Interestingly, FreeBSD does what Linux does not: after an fsync() failure, pages remain dirty, and subsequent fsync() calls will retry the write. This behavior has existed since 1999, proving that "keeping pages dirty" is technically feasible. However, FreeBSD also compromises -- if the device disappears entirely, newer versions will eventually give up on those pages.

https://medium.com/@baotiao/starting-from-postgresqls-fsync-failure-840af156585c