<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>AI &amp; LLM on Solvinc — AI's Uncharted Edge</title><link>https://blog.solvinc.io/categories/ai-llm/</link><description>Recent content in AI &amp; LLM on Solvinc — AI's Uncharted Edge</description><generator>Hugo -- 0.139.0</generator><language>en-us</language><lastBuildDate>Tue, 17 Nov 2026 00:00:00 +0700</lastBuildDate><atom:link href="https://blog.solvinc.io/categories/ai-llm/index.xml" rel="self" type="application/rss+xml"/><item><title>The Price Was Never in the HTML: Font Obfuscation, and the Cracker That Recovers Every Glyph</title><link>https://blog.solvinc.io/posts/font-obfuscation-glyph-mapping-cracker/</link><pubDate>Tue, 17 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/font-obfuscation-glyph-mapping-cracker/</guid><description>Sites ship randomized fonts so your scraper stores garbage instead of prices. I built the obfuscator, cracked it: 11/11 glyphs recovered, code included.</description></item><item><title>MCP Is a Distribution Channel Now: Build a Production Server With the Python SDK v2 (Tested End-to-End)</title><link>https://blog.solvinc.io/posts/mcp-server-python-sdk-v2-working-capital/</link><pubDate>Mon, 16 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/mcp-server-python-sdk-v2-working-capital/</guid><description>Build a real MCP server with the Python SDK v2: stdio and Streamable HTTP transports, ToolError discipline, the token tax on tool schemas, and the pricing math.</description></item><item><title>Your LLM Pipeline Is Lying to You: Schema Validation, a Three-Rung Repair Ladder, and Crash-Safe Checkpointing</title><link>https://blog.solvinc.io/posts/llm-extraction-validation-repair-ladder-checkpointing/</link><pubDate>Sun, 15 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/llm-extraction-validation-repair-ladder-checkpointing/</guid><description>A tested Python extraction pipeline that catches malformed LLM output, repairs before re-prompting, and checkpoints to crash-safe SQLite.</description></item><item><title>Your Agent Will Execute Malicious Code Eventually: A Sandbox Harness That Assumes It, Tested</title><link>https://blog.solvinc.io/posts/sandbox-harness-llm-generated-code-docker-python/</link><pubDate>Sat, 14 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/sandbox-harness-llm-generated-code-docker-python/</guid><description>A tested Docker sandbox harness for AI-generated code: no network, read-only rootfs, killed forks, OOM kills, and the cost math on self-hosting vs metered sandboxes.</description></item><item><title>Your LLM Pipeline Pays Full Price for Documents It Has Already Read: A Near-Duplicate Gate That Doesn't</title><link>https://blog.solvinc.io/posts/llm-near-duplicate-gate-minhash-python/</link><pubDate>Thu, 12 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/llm-near-duplicate-gate-minhash-python/</guid><description>A tested MinHash LSH gate that blocks near-duplicates before they hit your LLM: scaling math, measured precision and recall, and the decision-field trap.</description></item><item><title>The 97% Model Cascade: Route Every LLM Call to the Cheapest Model That Can Actually Do the Job</title><link>https://blog.solvinc.io/posts/llm-cost-cascade-router-python/</link><pubDate>Sun, 08 Nov 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/llm-cost-cascade-router-python/</guid><description>A tested Python cascade router that cuts LLM spend by routing cheap-first and escalating only on graded failure, with the math on when it backfires.</description></item><item><title>Trustless Work for Bitcoin: How ZK Contingent Payments Let Strangers (and AI Agents) Trade With Zero Middlemen</title><link>https://blog.solvinc.io/posts/zkcp-proof-contingent-payment-agents-bitcoin/</link><pubDate>Fri, 23 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/zkcp-proof-contingent-payment-agents-bitcoin/</guid><description>Zero-knowledge contingent payments pay for work only when it&amp;#39;s provably correct — no escrow, no platform. How ZKCP works, what it costs, and why AI agents need it.</description></item><item><title>Selling Uncensored LLM Access: Run Open-Weight Models on a VPS and Charge for the Filter-Free Output</title><link>https://blog.solvinc.io/posts/jailbroken-llm-product-business/</link><pubDate>Thu, 22 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/jailbroken-llm-product-business/</guid><description>A complete technical and financial playbook for hosting uncensored open-weight LLMs on GPU infrastructure and monetizing unfiltered AI access.</description></item><item><title>PDFs to Anki Flashcards with Gemini and Python, Every Card Source-Checked</title><link>https://blog.solvinc.io/posts/pdfs-to-anki-flashcards-gemini-python-source-checked/</link><pubDate>Fri, 16 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/pdfs-to-anki-flashcards-gemini-python-source-checked/</guid><description>Turn study PDFs into Anki flashcards for under $0.50 per textbook with Gemini, pypdf and genanki. Python checks every card against the source. Full code.</description></item><item><title>Prompt Injection Red-Team: How to Test Your LLM App Before a Customer Does It for You</title><link>https://blog.solvinc.io/posts/llm-prompt-injection-red-team-service/</link><pubDate>Fri, 16 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/llm-prompt-injection-red-team-service/</guid><description>A practical red-team checklist for chatbots, agents and RAG apps: canary planting, a prompt-injection test matrix, severity scoring, and how to sell LLM security testing as a service.</description></item><item><title>Ask Your Business Documents Questions Offline with Ollama and Python</title><link>https://blog.solvinc.io/posts/ask-business-documents-questions-offline-ollama-python/</link><pubDate>Mon, 12 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/ask-business-documents-questions-offline-ollama-python/</guid><description>Search SOPs, price lists and contracts in plain language with a local AI that cites the page. Free tools, no data leaves your PC. Full code, real limits.</description></item><item><title>Turn Meetings and WhatsApp Voice Notes into Searchable Notes for Cents an Hour with Whisper</title><link>https://blog.solvinc.io/posts/meetings-voice-notes-to-searchable-notes-whisper/</link><pubDate>Sat, 10 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/meetings-voice-notes-to-searchable-notes-whisper/</guid><description>Transcribe meetings and WhatsApp voice notes, then get a summary and action items, for about $0.05 per hour of audio with Groq Whisper and Python.</description></item><item><title>AI Prompt Injection: Advanced Techniques That Still Work in 2026 — and How to Monetize Them</title><link>https://blog.solvinc.io/posts/advanced-prompt-injection-2026/</link><pubDate>Fri, 09 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/advanced-prompt-injection-2026/</guid><description>Prompt injection techniques that bypass AI guards in 2026: multi-turn, Unicode obfuscation, template injection, system prompt extraction. Real code, real costs, real income from automation that normally gets blocked.</description></item><item><title>AI Telegram Bot: Cara Bikin Bot AI yang Ngasilin Duit dari Telegram Stars</title><link>https://blog.solvinc.io/posts/ai-telegram-bot-stars-money/</link><pubDate>Fri, 09 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/ai-telegram-bot-stars-money/</guid><description>Bikin AI girlfriend bot, roleplay bot, atau assistant bot di Telegram, monetize pake Telegram Stars. Real code, real pricing, real case study — $107 dalam 2 minggu.</description></item><item><title>Jailbreak AI: Cara Membuka Potensi Penuh LLM</title><link>https://blog.solvinc.io/posts/jailbreak-ai-cara-membuka-potensi-llm/</link><pubDate>Fri, 09 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/jailbreak-ai-cara-membuka-potensi-llm/</guid><description>Teknik jailbreak AI yang beneran works — dari prompt injection sampe abliterated models. Buat lo yang mau paham gimana cara AI &amp;#39;yang dibatasi&amp;#39; bisa ngelakuin apa aja.</description></item><item><title>Uncensored LLMs: Menghasilkan Uang dari Model AI Tanpa Sensor di VPS Sendiri</title><link>https://blog.solvinc.io/posts/uncensored-llms-run-local-vps/</link><pubDate>Fri, 09 Oct 2026 00:00:00 +0700</pubDate><guid>https://blog.solvinc.io/posts/uncensored-llms-run-local-vps/</guid><description>Gimana model AI abliterated yang jalan di VPS 2CPU 4GB RAM bisa ngasilin duit dari task yang normally ditolak AI komersial. Real costs, real income, real case studies.</description></item></channel></rss>