Saurabh Prakash
@saurabhp.bsky.social
I dont know https://www.codeandcontext.dev
Effect.ts doesn't click until the mental models do — generics, ADTs, composition, laziness. Broke them down from first principles. Link in comments. www.codeandcontext.dev/blog/effect-... #EffectTS #TypeScript #FunctionalProgramming #SoftwareArchitecture #Programming #FP
Effect.ts: The Prerequisites You Need Before the Library Clicks
Before Effect.ts feels natural, you need four mental models: generics, algebraic data types, functional composition, and laziness. This first post builds those foundations from first principles.
codeandcontext.dev
AI is no longer just writing code. It’s starting to break it too. Anthropic’s Project Glasswing is a strong signal that cybersecurity is entering a new phase, where both vulnerability discovery and defense become AI-accelerated. Details in thr post: www.codeandcontext.dev/blog/project...
Project Glasswing: When AI Becomes Both the Threat and the Defense
Anthropic's Project Glasswing signals a new cybersecurity era: frontier models can now discover and exploit vulnerabilities at machine speed, forcing industry to build defense before broad deployment.
codeandcontext.dev
What if the deepest insight in AI is also the oldest one in Advaita philosophy? Everything you think, feel, and believe about yourself is inference — tokens burning compute. But the awareness behind it all? Zero parameters. Always on. Never computed. 👇 www.codeandcontext.dev/blog/advaita...
Advaita for AI Practitioners
A contemplative mapping between Advaita concepts and AI inference: everything is inference except your own being.
codeandcontext.dev
Most LLM apps are being secured like chatbots. But the real attack surface is much bigger: The OWASP Top 10 for LLM Apps 2025 makes that painfully clear. I broke down all 10 risks in plain technical detail, and what they mean for real AI systems. Read it here: www.codeandcontext.dev/blog/owasp-t...
OWASP Top 10 for LLM Applications 2025: A Technical Breakdown
The 2025 update to the OWASP Top 10 for LLM Applications reflects the rise of agentic frameworks, RAG pipelines, and multi-modal AI. From prompt injection to unbounded consumption, here is a detailed ...
codeandcontext.dev
Everyone's waiting for an AGI announcement. Press release. Benchmark. Big moment. I think we already crossed it — we just didn't notice. I wrote about why the AGI age may have started without an announcement. Link in the comments 👇 www.codeandcontext.dev/blog/we-alre...
We Already Crossed AGI. You Just Didn't Notice.
Everyone is waiting for the AGI announcement. But the real shift already happened — reasoning became infrastructure, thinking stopped being scarce, and the token became its currency.
codeandcontext.dev
Just published: WebMCP Explained AI browser automation is still too brittle. Agents inspect pixels, scrape DOMs, and break when the UI changes. WebMCP proposes a better model: websites expose structured tools directly to agents. Read here: www.codeandcontext.dev/blog/webmcp-...
WebMCP Explained: The Web Standard Proposal That Could Fix AI Browser Automation
A technical breakdown of WebMCP, the browser-based proposal for AI agents that replaces brittle UI automation with structured tools, typed inputs, and human-in-the-loop workflows.
codeandcontext.dev
Chrome 146 is a real unlock for web agents With one toggle, your live browsing session becomes available over MCP. I wrote about that shift here: www.codeandcontext.dev/blog/web-aut...
Web Automation for AI Agents: Playwright, Agent Browsers, and the Rise of WebMCP
A practical comparison of three ways AI agents can use the web: DOM-level automation with Playwright, visual browser agents, and an emerging MCP-inspired model for agent-native websites.
codeandcontext.dev
Most AI agent builders are using frameworks without understanding the 7 papers that shaped reasoning, planning, tool use, memory, and RAG. I broke them down in one post so you can understand how modern agents actually work: www.codeandcontext.dev/blog/seven-r...
The 7 Research Papers That Explain How Every AI Agent Actually Works
A layered reading path through the seven papers that explain the reasoning, planning, tool use, memory, and grounding behind modern AI agents.
codeandcontext.dev
AI is making knowledge work cheaper—and that’s exactly why we’ll end up doing more of it. In my latest post, I break down 4 counter-intuitive reasons this happens (through the lens of Jevons Paradox): Efficiency creates new demand, not less demand. www.codeandcontext.dev/blog/ai-para...
The AI Paradox: 4 Reasons Why Making Work Cheaper Creates More of It
History shows efficiency doesn't reduce work—it explodes it. Here's why AI will create more opportunities, not fewer, through the lens of Jevons Paradox.
codeandcontext.dev
Why does Elasticsearch default to k1=1.2 and b=0.75? I wrote the post I wish existed when I started building search systems From inverted indexes → TF-IDF → BM25 → Hybrid RAG pipelines. Every step motivated by a concrete failure in the previous one www.codeandcontext.dev/blog/from-in...
From Inverted Indexes to BM25: How Search Ranking Actually Works
A first-principles walkthrough of search — from scanning documents to inverted indexes, TF-IDF to BM25, and why modern systems combine lexical and semantic search.
codeandcontext.dev
Karpathy open-sourced "autoresearch" — an AI agent that optimizes ML models while you sleep. The loop: edit training code → run a 5-min experiment → keep if the metric improves → repeat. No human needed.
Autoresearch: 700 Experiments While You Sleep
An AI agent ran 700 training experiments autonomously and found 20 real improvements. Here's how autoresearch works.
codeandcontext.dev
Knuth on Claude solving his open problem: "Shock! Shock!" When Knuth updates on AI, pay attention. Karpathy. Torvalds. Hashimoto. All noticing the same shift. Not faster code. AI executing engineering work. My take 👇
When Knuth Says 'Shock!': Why Agentic Coding Is a Real Inflection Point
Knuth, Karpathy, Torvalds, and Hashimoto are converging on the same signal: coding is shifting from manual implementation to agent orchestration, verification, and taste.
codeandcontext.dev
Written in 2019. Sat in my drafts for years. Wish the world hadn't made it relevant — but here we are. Enjoy. www.codeandcontext.dev/blog/carrier...
Carrier Dialogue in the modern times
A fictional midnight dialogue between a father-and-son mosquito duo, reflecting on knowledge, intelligence, survival, and irony.
codeandcontext.dev
Are you a thinker or a token churner? "Agents turning into thinkers, I found, Seeing thinkers, churning tokens, my wonder has no bound, Hey, you are not a thinker or a token churner, but a being from beyond, no matter how I sound Aquarian age has arrived, murmurs are doing round."
