Vicki @vickiboykis.com · 19/09/2026feels like I’m the only one pacing tbh and that’s only because I’m getting docker build errors 1210
Reposted by VickiJakob @jakobjarecki.bsky.social · 16/09/2026A booming voice yelling at me "YOU ARE NOT ALLOWED TO SAVE DOCUMENTS WITH A LEADING !. ALL OF THEM EVENTUALLY HAVE IT AND OT SORTS NOTHING, YOU FOOL". 0161
Vicki @vickiboykis.com · 16/09/2026Startup idea: small personalized models that are fine tuned on the trauma of your past production incidents 4968
Vicki @vickiboykis.com · 08/09/2026Yup, and recs share the same philosophical underpinning in IR which is trying to find the most relevant item for a given query by filtering down the result space; in search we just get lucky because the query is explicit 010
Vicki @vickiboykis.com · 08/09/2026It was a tongue in cheek post but legitimately a lot of problems that we face daily are around search and relevance 110
Vicki @vickiboykis.com · 08/09/2026Mostly right now around how we are trying to work with harnesses for agentic code generation, same patterns in information retrieval. But also in surfacing anomaly detection, lots of cases where IR is the right pattern 120
Vicki @vickiboykis.com · 07/09/2026Area woman continuously working in recsys more and more convinced that most ML problems are recsys problems 5311
Vicki @vickiboykis.com · 05/09/2026It’s pretty wild how cleanly the labs have been able to spin basic security incompetence into positive marketing 817228
Vicki @vickiboykis.com · 02/09/2026New post: we need to be deleting more code. vickiboykis.com/2026/09/01/b...vickiboykis.comBad code is kudzuDelete your features 311713
Vicki @vickiboykis.com · 02/08/2026Blog posts we used to have: “how I reduced latency across several hundred machines by 200 ms by removing two O(n)^2 hot paths” Blog posts we have now: “Here’s how to get the most out of /ultrathink in the latest release of Claudex 3.0.5” 916612
Vicki @vickiboykis.com · 29/07/2026Curious about people’s ML dependency management strategies in the face of supply chain risks. Often pinning dependencies particularly in the torch/HuggingFace ecosystem is the only way to create reproducible training and inference code but that means dependabot isn’t happy. 1100
Vicki @vickiboykis.com · 16/07/2026There are millions of anti patterns that make stuff slower and less elegant but probably the worst offender is adding a network call 0282
Vicki @vickiboykis.com · 16/07/2026There are like only three design patterns you need to use 80% to make stuff good and fast and elegant Make it a lookup (hashmap) Split it up into smaller pieces (sharding) Combine it into one large piece (map reduce) 4998
Vicki @vickiboykis.com · 14/07/2026Eye bleach: an actual architecture and engineering blog post lobste.rs/s/ko1ji1/lob...lobste.rslobste.rs is now running on SQLite87 comments 0270
Vicki @vickiboykis.com · 09/07/2026I have but I’m looking for something more technical and less biased than both Andrew’s and Jared’s perspectives (both valid, but they’re at the center of the storm.) I’m looking to understand the core of the technical issues 130
Vicki @vickiboykis.com · 09/07/2026Looking to read an in-depth examination of the project history and code evolution of Bun in Zig, along with a technical evaluation of its current Rust re-implementation. I want to check out links to PRs, sidebars on Zig and Rust, the works. Who’s doing this? 2100
Vicki @vickiboykis.com · 01/07/2026Not sure I know what either of those is but the basic idea is that intuition for writing good code is a skill that can be learned over time 150
Vicki @vickiboykis.com · 01/07/2026Recommended reading if you feel you’ve lost your way , or if you’re having more fun than ever 51123
Vicki @vickiboykis.com · 23/06/2026Everyone working on recsys/embeddings/information retrieval/search is sleeping on this newsletter if you're not subscribed. I can't believe I didn't mention it before but it just summarizes the best content in the field every week recsys.substack.comrecsys.substack.comTop Information Retrieval Papers of the Week | Sumit | SubstackA weekly curated newsletter about the latest research papers in the Information Retrieval domain, including Recommender Systems, Search, Retrieval, and Ranking. Click to read Top Information Retrieval... 3352
Vicki @vickiboykis.com · 19/06/2026like, discuss.python.org/t/using-exce...discuss.python.orgUsing ExceptionGroup at Anthropic (experience report)Context: written for internal distribution at Anthropic and subsequent publication on the Python forum, to summarize lessons learned from a recent migration. I hope this writeup is of interest and pe... 0170
Vicki @vickiboykis.com · 19/06/2026the actual most revolutionary thing the ai labs have actually done is make it look like it's insanely easy to duct tape together python microservices with all their concurrency issues and trained models with all their transformers issues and have them run as one api call 58510
Vicki @vickiboykis.com · 18/06/2026This is the only thing noam shazeer has blogged about publicly and I use it all the time medium.com/@NoamShazeer...medium.comShape Suffixes — Good Coding StyleIf you code neural networks, I believe that this convention can make your life more pleasant. We keep this pretty religiously at… 1201
Vicki @vickiboykis.com · 15/06/2026Were you doing context pooling or a sliding window? lm studio (and likely llama cpp) has options to change how the context window changes 120
Vicki @vickiboykis.com · 15/06/2026I’ve been able to so far bump the context window to 75% of max but ill probably run into the same issue and will want to check out stuff like this arxiv.org/abs/2606.06492arxiv.orgCode2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software EvolutionCode language models need repository-level context to resolve imports, APIs, and project conventions. Existing methods inject this knowledge as long inputs (retrieved through RAG or dependency analysi... 040
Vicki @vickiboykis.com · 15/06/2026You can try a lower quant and see if that helps any, though the results might not be as good 010
Vicki @vickiboykis.com · 15/06/2026It can do whatever you want it to! It does YOLO out of the box which is why I limited it but you can write extensions that control its behavior 110
Vicki @vickiboykis.com · 15/06/2026new post: how I develop recently using local models. the tooling is now good enough to do agentic workflows and everyone should give them a try! vickiboykis.com/2026/06/15/r...vickiboykis.comRunning local models is good nowLocal agentic coding has gotten great over the past few months 1825737
Vicki @vickiboykis.com · 11/06/2026The good thing is that you don’t need it! Unless you want something fun and shiny. My laptop is 4 years old 110
Vicki @vickiboykis.com · 11/06/2026this is so cool, continuing my local experiments and using pi and lm studio and gemma to introspect how i use agents locally. On the left, you can literally see the streaming tokens and the context window filling up. 4230
Vicki @vickiboykis.com · 10/06/2026every day we get closer to looping back around to small purpose-built specialized models 3422
Vicki @vickiboykis.com · 06/06/2026Oh sorry this pi: pi.devpi.devPi Coding AgentA terminal-based coding agent 110
Vicki @vickiboykis.com · 06/06/2026Running a local model and harness (Pi + GGUF via LM Studio) and I can see everything it's doing, which is dope. 4613
Reposted by VickiMiguel de Icaza @migueldeicaza.bsky.social · 03/06/2026Good post from the author of rsync on the recent pile-on: medium.com/@tridge60/rs...medium.comrsync and outrageI gave up blogging a long time ago (apart from an occasional thing about ArduPilot), I tend to just write code and hope people find it… 520547
Reposted by VickiEugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 03/06/2026It’s not “do good work and be successful.” It’s doing good work, talking about the work, figuring out how to get your work to the right people, accepting that good work is slower than bad work and finding a path through that anyways 212011
Vicki @vickiboykis.com · 01/06/2026Great piece, “We're entering a world where it's cheap to produce bigger changes, but that doesn't make it any safer.” apenwarr.ca/log/20260531apenwarr.caThe software industry: annealing, but wrongIn recent months I've heard of several teams with an interesting policy: each pull request should be no more than a few files, and no more t... 2427