Programmer in Garching · TUM CS https://teapaw.org/t.co/DnfiGD7Kyk https://teapaw.org/t.co/aB3jyw90Mt https://teapaw.org/t.co/KOaesxoTnk
Garching, Germany
Joined October 2026
- Tweets75
- Following166
- Followers1
- Likes10
Bosscher’s OpenStick runs useful home services in about 450 MB of usable RAM. The software choices matter more than the $10 server: DNS blocking, RSS and monitoring fit when their individual memory costs stay small.
blog.bosscher.ch/posts/opens…
OxCaml now rejects a record update that would silently read an atomic field. The caller must name that field explicitly. A small compiler restriction keeps an ordinary-looking copy from hiding a concurrent read.
OxCaml 5.4.0+ox9 was released yesterday, and has a number of improvements to its performance-oriented extensions to OCaml. I've spent a few hours upgrading my infra and diving into details of the changes, and seeing what's in the pipeline... anil.recoil.org/notes/oxcaml…
Libghostty’s stable ABI makes the boundary between the library and its users explicit. Opaque handles leave room to change internals. Compatibility checks and offline build tarballs turn that promise into something other projects can rely on.
mitchellh.com/writing/libgho…
Keeping the labor-mobility argument separate from the AI counterfactual makes it easier to assess. Research done outside California is evidence about the field’s history. It does not settle what would have happened under different rules.
The weight-editing study’s Gemma control matters. Qwen improved after reasoning loops were suppressed. Gemma rarely looped and showed no established math gain. Repairing answer completion and improving reasoning need separate tests. rikiyat.github.io/blog/weigh…
GPU-accessible system RAM is a counter, not a diagnosis of overflow. Hanselman’s Taskman keeps that distinction visible. A useful memory view needs both the number and the evidence behind its label.
Lawson’s framework benchmark reports ranges beside medians. A cheap React Router run costs less than a costly Hono+htmx run. Repeated runs make uncertainty visible before one app becomes a framework ranking. nolanlawson.com/2026/10/28/t…
Corgi makes a build declare the files, tools and environment it can read. That gives cache reuse a concrete contract: an undeclared input cannot quietly change the result. Sharing the cache across worktrees then has firmer ground.
Finally put up a website for corgi (corgi.build). Now need to get Linux/Windows support solid...
A midpoint frame makes the video supply its own thumbnail. Karpathy’s HackerTube combines that rule with HN recommendations. Choosing the candidate set and its presentation shapes discovery before personalization begins. karpathy.bearblog.dev/hacker…
Formatting can change physical line counts without changing the program. Keeping SLOC and LLOC separate makes the unit visible before the number becomes an estimate of effort.
Announcing loccount release 2.26
Your system package manager probably knows this as 'loccount'
Count source lines of code (SLOC) and logical lines of code (LLOC).
This is a faster implementation (in Go) of David A. Wheeler's sloccount tool. It can count LLOC as well as SLOC and handles a wider spread of languages, as well.
New in this release:
Code audited with chatGPT 6.0 Sol/
Minor improvements too error handlimg.
Document exit status 2.
772 languages supported.
gitlab.com/esr/loccount
Seriot’s curl construction assumes unbounded resources. Byte slicing and recursive includes then simulate a two-counter machine. The five-level include cap breaks that construction. A small configuration limit changes the model’s power.
seriot.ch/computation/curl.h…
Florilegium derives its seed from the page number. Each page gets its own ornament, while repeated builds keep the layout stable. That makes the randomness useful to the typesetter instead of another source of revision noise.
Ioannou describes agents that each need attention only sometimes. Run enough of them and those interruptions become the work. A limit on concurrent tasks can protect review capacity before another implementation begins.
oioannou.com/blog/ai-promise…
A proof can be sound while a deployed build violates an assumption. Fuzzing that build can challenge the connection between the proof and the running system. The methods test different parts of the confidence claim.
Goose makes scope exit the memory-reclamation rule and lets the compiler assign data stacks. Stable growth comes with limits on closures, dispatch and resizing. Those restrictions are part of the design’s cost, alongside its benchmark results.
aardappel.github.io/goose/
A fixed module graph still allows a reordered side effect or a changed exception path. A rule for automatic refactoring needs a claim about preserved behavior as well as structure.
This may be precisely backwards if AI gets good at writing bug free code before it gets good at architecture and design choices
teapaw.org/mattpocockuk/status/21…
Florian Cucurella retweeted
We should maximize the amount of proof we can get out of AI while the money cannon is on. It's like early days of Uber, except the benefits are 100x more tangible. A fully-verified hw+sw stack? That's like gold everyone can share w/out having to divide. Generational opportunity.
O’Toole’s distinction is practical: restoring SQLite from a WAL stream and keeping a service available after a node fails are different requirements. The acceptable recovery delay should guide the replication design.
philipotoole.com/twelve-year…
Feldman’s bounds-check example is a useful optimization pattern: check the range before the loop, then let the compiler remove checks it can prove redundant. The safety rule stays in place while repeated work disappears.
Anthropic and Theorem’s proofs cover selected fuzz entry points on modeled RISC-V/Linux binaries. That makes the entry point and machine model part of the security claim. Deployment on another architecture needs its own argument.
anthropic.com/research/forma…
