https://by-sal.dev DEV BY SAL is a small senior software studio that builds AI-native products, frontier infrastructure and revenue-shifting platforms for ambitious companies. en-us Wed, 20 May 2026 05:07:53 GMT https://by-sal.dev/insights/the-ai-product-engineer https://by-sal.dev/insights/the-ai-product-engineer Mon, 18 May 2026 00:00:00 GMT Sal Anvarov Insights Senior product engineering used to mean comfort across the front-end, the back-end, and the database. The bar has moved. The teams shipping consequential software in 2026 demand a fourth competency — turning probabilistic models into reliable product behavior — and the engineers who have it are the highest-leverage hires on the org chart. https://by-sal.dev/blog/the-four-week-mvp https://by-sal.dev/blog/the-four-week-mvp Thu, 14 May 2026 00:00:00 GMT Sal Anvarov Blog People keep asking why we ship in four weeks instead of six months. It's not because we're faster. It's because four weeks is what's actually required to learn anything. https://by-sal.dev/papers/eval-driven-development-for-llm-systems https://by-sal.dev/papers/eval-driven-development-for-llm-systems Tue, 12 May 2026 00:00:00 GMT Sal Anvarov Paper Most production LLM systems are evaluated by other LLMs, against benchmarks that have probably leaked into the model's training data, with judges that systematically prefer their own kin. The recent literature is clear-eyed about all three problems. This paper is a working playbook for shipping evaluation infrastructure that survives those facts. https://by-sal.dev/papers/agent-architecture-patterns-2026 https://by-sal.dev/papers/agent-architecture-patterns-2026 Fri, 08 May 2026 00:00:00 GMT Sal Anvarov Paper The literature on production LLM agents has matured from "can we make this work at all?" to "which interface and memory structure makes it work reliably?" This paper walks the dominant patterns — ReAct, Reflexion, AutoGen-style multi-agent, and SWE-agent's interface-as-product — and the failure modes that come with them. https://by-sal.dev/insights/the-cost-of-the-second-pilot https://by-sal.dev/insights/the-cost-of-the-second-pilot Mon, 04 May 2026 00:00:00 GMT Sal Anvarov Insights AI pilots are easier to start than ever and harder to ship than they look. The teams that fail almost all fail in the same place — between the first design partner and the second customer, when the demo has to become a product. This is what that gap actually costs, and how to budget for it. https://by-sal.dev/blog/why-we-dont-use-llm-frameworks https://by-sal.dev/blog/why-we-dont-use-llm-frameworks Sat, 02 May 2026 00:00:00 GMT Sal Anvarov Blog LangChain, LlamaIndex, Haystack — we have used them all and we ship most of our agents without them. Here is when frameworks help, when they hurt, and why we usually choose the harder-looking option. https://by-sal.dev/insights/why-build-versus-buy-just-flipped https://by-sal.dev/insights/why-build-versus-buy-just-flipped Tue, 21 Apr 2026 00:00:00 GMT Sal Anvarov Insights For two decades, "buy the SaaS" was the right answer for almost any internal-tool question. The economics that made that true are no longer true. We walk through what changed, what it means for vendor selection, and which categories are most exposed. https://by-sal.dev/blog/the-eval-set-is-the-spec https://by-sal.dev/blog/the-eval-set-is-the-spec Sat, 18 Apr 2026 00:00:00 GMT Sal Anvarov Blog For LLM-shaped products, the document that defines what the product does is the eval set, not the PRD. We have stopped writing PRDs for AI features and started writing the cases. The work is better for it. https://by-sal.dev/papers/rag-that-actually-retrieves https://by-sal.dev/papers/rag-that-actually-retrieves Sun, 15 Mar 2026 00:00:00 GMT Sal Anvarov Paper Production RAG in 2026 looks nothing like the "embed, retrieve, generate" of 2023. The recent literature converges on retrieval as a routing problem — when to retrieve, how to verify retrieval quality, when to switch to graph indexing, when to lean on long-context instead. This paper is the practitioner-facing version.