Orçun Başlak

Yapay zeka'yı efektif kullanabilenler ve kullanamayanlar diye 2 grup çok hızlı oluşacak. Kullanamayanları zorlu günler bekliyor.

Bi evimizin önüne rüzgar türbini yapılmamıştı :)

@karpathy I have a similar setup but I wasn't really comfortable with Obsidian esp. with the terminal integration. I found myself switching back and forth with vscode.

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

@memetsimsek Vergi risklerini tespit etmek icin beyan oncesinde malikontrol.com adresinden risklerini analiz edebilirsiniz.

Gelir vergisi beyan döneminin sonuna yaklaşılırken, mükelleflerimizi beyannamelerini eksiksiz ve zamanında vermeye davet ediyoruz. Yapay zekâ destekli çalışmalarla kayıt dışı faaliyetlere ilişkin tespit ve denetim kapasitemizi önemli ölçüde artırıyoruz. Vergi adaleti ve kamu maliyesinin sürdürülebilirliği için gönüllü uyum büyük önem taşıyor.

@borsacibaba Malikontrol.com adresinden sirketinizi denetlettirebilirsiniz. Sorun cikmadan once haberdar olabilirsiniz.

Artık vergi kaçırmak neredeyse imkansız hale geldi Vergi Dairesi, 3 yapay zeka destekli sistemle sahte faturaları, birinci derece yakınlara yapılan IBAN transferlerini ve 10.000 TL’ye kadar işlemleri saniyeler içinde tarayabiliyor

@fatih I tried Ghostty but it never gave the stability i have in iTerm2. I don't the hype with terminals nowadays. What makes u stick to it?

I tried Codex’s macOS app. It’s great but nothing can beat yet my Tmux + CLI Agent + Neovim setup. I can deploy, debug on to go with tmux, while also interact with the Agent. Codex has a terminal, but it’s not exactly like Ghostty or my Tmux setup. Sometimes I want to open a file, read or jump from one place to another. Codex can’t do that either. So I don’t know how the answer look like. I think IDE’s with Agent support also are limited. For example I love Cursor for planning/building large projects. But it doesn’t quite feel like how I would use my terminal setup. The VSCode UI is also limiting, leaking into Cursor product. In my Terminal, the user interface is fluid. I open a new tab, and I have a blank scratch. I can open two agents side by side, or stream logs from a remote server while debugging it with an Agent. Or have a 3-window vertical Neovim setup. Whatever the solution is, if it’s not the CLI, it needs to be accommodating various people needs. Or it has to be quite different in design and bring a new paradigm.

@sama We still need a "plan mode" Sam. I don't feel comfortable proceeding without knowing what the agent thinks.

From how the team operates, I always thought Codex would eventually win. But I am pleasantly surprised to see it happening so quickly. Thank you to all the builders; you inspire us to work even harder.

NodeJS ile typescript'in backend'e geçişini çok garipsemiştim. Şuanda typescript ile CLI yazıldığını görmek o kadar şaşırtıcı ki. Demek ki neymiş? Userbase herşeymiş. Çok ihtiyaç olan yerde C/Rust binding; geri kalanı TS devam.

Meteorologlar kapatılsın.

Synology Photos çok kötü ama Synology docket destekliyor. Docker compose ile Immich kurdum. Çok mutluyum.

Bye bye Google One and Google Photos. Welcome Synology NAS.

Yeni iOS ana ekran tasarımım. Çok temiz oldu. Sağ alt baş parmak ile kullanmaya çok uygun.

@erenbali The old logo was much better.

I don't wanna be the guy who hates new logos but if I saw this on the street I'd think it's the women's bathroom

@KorayBahar @BlindLook @figopara Late nite DJ voice gibi olmus :)

@BlindLook @figopara Link Burada: blindlook.com/tr/blogs/koray…

Yerebatan Sarnıcı'nı Betimledim, Ses Oldum! Siz de dinlemek ve parçası olmak isterseniz sizi buraya bekliyoruz. blindlook.com @BlindLook @figopara

@fatih Girişimci olursan işler değişir.

ABD’de değer üreten bir suru oluşumda yer aldım, Pazar günü çalışan görmedim. On-call durumu varsa varını yoğunu koyarsın. Bazen önemli bir ürün çıkar, yine belirli dönemlerde yardırırsın. Bunların dışında bunu güzel göstermenin bir anlamı yok. Haftasonları ve boş günleriniz kendiniz içindir.

Aramalarımın neredeyse %90'ı ChatGPT üzerine kaydı. Bence artık web sitelerini de LLM compatible hale getirmeliyiz. XML site index gibi artik LLM index konusu hayatımıza girmeli.

Hoşgeldin #bitchat 🚀

Almanya'da fişlerin altında şifreleme var. Sistemi merkezileştirmeden nasıl çözebiliriz sorusuna güzel ve lokal/dağıtık bir çözüm olmuş. Diğer uçta bizim eFatura sistemi; tüm sistem merkezi çalışıyor. İyi düşünülmüş.

Feminizm artık kadınları özgürleştiriyor mu, yoksa onları yeni bir kalıba mı sokuyor? orcun.baslak.com/feminizm-uzeri…

@fatih Supabase hakkında ne düşünüyorsun?