<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>bitvaria GmbH Blog</title><description>Posts on AI applications for mid-sized companies: cost, architecture and operations.</description><link>https://bitvaria.com/</link><language>en</language><item><title>The Quality Gap in Open AI Models, Honestly Quantified</title><link>https://bitvaria.com/en/qualitaetsluecke-offener-modelle/</link><guid isPermaLink="true">https://bitvaria.com/en/qualitaetsluecke-offener-modelle/</guid><description>In June an open model cracked 50 points on the leading independent AI benchmark for the first time; six weeks later a second one came within three points of the frontier. Why the average number is still almost irrelevant for your own decision.</description><pubDate>Fri, 14 Aug 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>model-selection</category><category>open-models</category><category>benchmarks</category></item><item><title>Why the Same Model Almost Never Gives the Same Answer Twice</title><link>https://bitvaria.com/en/sampling/</link><guid isPermaLink="true">https://bitvaria.com/en/sampling/</guid><description>Same prompt, same model, twice, two different answers. Not a bug: the model draws from a probability distribution at every word. What temperature, top-p, and top-k actually set.</description><pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate><category>ki</category><category>ai-engineering</category><category>sampling</category><category>temperature</category><category>top-p</category><category>llm</category></item><item><title>What Happens When Your AI Vendor Disappears</title><link>https://bitvaria.com/en/was-passiert-wenn-der-dienstleister-wegfaellt/</link><guid isPermaLink="true">https://bitvaria.com/en/was-passiert-wenn-der-dienstleister-wegfaellt/</guid><description>An AI vendor can disappear no matter how big it was. Three questions decide whether that means a migration for your company or a rebuild from zero, and they belong before the contract, not after.</description><pubDate>Tue, 11 Aug 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>lock-in</category><category>vendor-selection</category><category>sovereignty</category></item><item><title>Paid Per Seat, Used Per Task</title><link>https://bitvaria.com/en/bezahlt-pro-lizenz-genutzt-pro-aufgabe/</link><guid isPermaLink="true">https://bitvaria.com/en/bezahlt-pro-lizenz-genutzt-pro-aufgabe/</guid><description>300 paid Copilot seats are not 300 users. What Microsoft&apos;s own reports count as &quot;active,&quot; what the widely cited 20-30 percent figure actually measures, and why it likely understates the gap rather than overstates it.</description><pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>copilot</category><category>lizenzkosten</category><category>nutzungsmessung</category></item><item><title>Residency Is Not Jurisdiction</title><link>https://bitvaria.com/en/residenz-ist-nicht-jurisdiktion/</link><guid isPermaLink="true">https://bitvaria.com/en/residenz-ist-nicht-jurisdiktion/</guid><description>Data &quot;stored in the EU&quot; sounds like protection from US access. It is not, automatically. What the CLOUD Act and FISA 702 mean, and when it matters for your company.</description><pubDate>Thu, 23 Jul 2026 00:00:00 GMT</pubDate><category>ki</category><category>datenschutz</category><category>souveraenitaet</category><category>mittelstand</category><category>compliance</category></item><item><title>Why the Same AI Request Has Two Different Speeds</title><link>https://bitvaria.com/en/prefill-decode/</link><guid isPermaLink="true">https://bitvaria.com/en/prefill-decode/</guid><description>One request answers instantly, then drips out word by word. Another takes a while before the first word appears, then runs through fast. The reason: two phases with different bottlenecks.</description><pubDate>Wed, 22 Jul 2026 00:00:00 GMT</pubDate><category>ki</category><category>ai-engineering</category><category>prefill-decode</category><category>inference</category><category>llm</category></item><item><title>How Does an AI See an Image? Why It Reads the Screenshot but Misses the Tiny Number</title><link>https://bitvaria.com/en/wie-sieht-ki-ein-bild/</link><guid isPermaLink="true">https://bitvaria.com/en/wie-sieht-ki-ein-bild/</guid><description>An AI model never sees an image as pixels. It cuts it into a grid of tiles that become numbers, and reads them like tokens. Why the fine print disappears first.</description><pubDate>Mon, 13 Jul 2026 00:00:00 GMT</pubDate><category>ki</category><category>ai-engineering</category><category>multimodalität</category><category>vision</category><category>vlm</category></item><item><title>The Threshold Where Custom Software Pays Off Has Dropped</title><link>https://bitvaria.com/en/wann-lohnt-sich-eigene-software/</link><guid isPermaLink="true">https://bitvaria.com/en/wann-lohnt-sich-eigene-software/</guid><description>Custom software for small problems used to be too expensive. In 2026 that math has shifted. What it means for internal tools in mid-sized companies, and where the line still sits.</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>ki</category><category>software</category><category>mittelstand</category><category>automatisierung</category><category>build-in-public</category></item><item><title>Hosting Your Own AI Models or Renting? The Honest Math</title><link>https://bitvaria.com/en/ki-modelle-hosten-oder-mieten/</link><guid isPermaLink="true">https://bitvaria.com/en/ki-modelle-hosten-oder-mieten/</guid><description>Is your own server for open AI models worth it, or is a subscription cheaper? The honest cost and hardware math for the Mittelstand, as of 2026.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate><category>ki</category><category>infrastruktur</category><category>kosten</category><category>open-source</category><category>mittelstand</category></item><item><title>Why a Model Gets More Expensive the Longer It Talks</title><link>https://bitvaria.com/en/kv-cache/</link><guid isPermaLink="true">https://bitvaria.com/en/kv-cache/</guid><description>Same model, same question. Every answer costs more the longer the conversation runs. The reason has a name: KV cache.</description><pubDate>Sat, 20 Jun 2026 00:00:00 GMT</pubDate><category>ki</category><category>ai-engineering</category><category>kv-cache</category><category>inference</category><category>llm</category></item><item><title>What Is a Token? Why a Language Model Sees No Text</title><link>https://bitvaria.com/en/was-ist-ein-token/</link><guid isPermaLink="true">https://bitvaria.com/en/was-ist-ein-token/</guid><description>A token is the smallest unit a language model splits text into. Why the model sees no letters, and what that one fact explains about cost and errors.</description><pubDate>Mon, 15 Jun 2026 00:00:00 GMT</pubDate><category>ki</category><category>ai-engineering</category><category>token</category><category>tokenizer</category><category>llm</category></item><item><title>What Does an AI Project Cost in the Mittelstand?</title><link>https://bitvaria.com/en/was-kostet-ki-projekt-mittelstand/</link><guid isPermaLink="true">https://bitvaria.com/en/was-kostet-ki-projekt-mittelstand/</guid><description>Custom AI applications for mid-sized companies cost between 15,000 and 60,000 euros. This article shows which factors determine the price and when the investment pays off.</description><pubDate>Tue, 24 Mar 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>kosten</category><category>ki-projekt</category></item><item><title>Why ChatGPT Does Not Know Your Company Knowledge</title><link>https://bitvaria.com/en/warum-chatgpt-ihr-unternehmenswissen-nicht-kennt/</link><guid isPermaLink="true">https://bitvaria.com/en/warum-chatgpt-ihr-unternehmenswissen-nicht-kennt/</guid><description>ChatGPT knows a lot about the world. About your company, it knows nothing. Here is how to make internal knowledge accessible to AI.</description><pubDate>Tue, 10 Feb 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>unternehmenswissen</category><category>rag</category><category>wissensmanagement</category></item><item><title>Custom AI Application vs. Off-the-Shelf Software: When Does Each Pay Off?</title><link>https://bitvaria.com/en/individuelle-ki-anwendung-vs-standardsoftware/</link><guid isPermaLink="true">https://bitvaria.com/en/individuelle-ki-anwendung-vs-standardsoftware/</guid><description>Off-the-shelf software covers 80 percent of cases. The remaining 20 percent decide your competitive advantage. Here is how to make the right call.</description><pubDate>Tue, 13 Jan 2026 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>standardsoftware</category><category>individuelle-entwicklung</category></item><item><title>Adopting AI Without an In-House IT Department</title><link>https://bitvaria.com/en/ki-einfuehren-ohne-it-abteilung/</link><guid isPermaLink="true">https://bitvaria.com/en/ki-einfuehren-ohne-it-abteilung/</guid><description>Most mid-sized companies have no ML specialists on staff. That is not an obstacle to AI projects, as long as the approach is right.</description><pubDate>Tue, 18 Nov 2025 00:00:00 GMT</pubDate><category>ki</category><category>mittelstand</category><category>it-abteilung</category><category>ki-einfuehrung</category></item><item><title>Stop Asking for Buttons: Modern IT Collaboration Done Right</title><link>https://bitvaria.com/en/working-with-your-it/</link><guid isPermaLink="true">https://bitvaria.com/en/working-with-your-it/</guid><description>Learn how modern IT collaboration thrives by focusing on problems, not just solutions. Discover practical strategies to empower IT teams and deliver smarter, business-aligned outcomes.</description><pubDate>Tue, 17 Dec 2024 00:00:00 GMT</pubDate><category>Collaboration</category><category>IT Strategy</category><category>Business Alignment</category><category>Digital Transformation</category><category>Project Success</category></item><item><title>Real-Time Alerts and Zero Data Loss: The AWS Lambda Pattern with Effective Monitoring You Can&apos;t Ignore</title><link>https://bitvaria.com/en/monitored-lambda-queue/</link><guid isPermaLink="true">https://bitvaria.com/en/monitored-lambda-queue/</guid><description>Discover how to create a reliable and monitored asynchronous processing system with AWS Lambda. This pattern combines SQS queues, dead letter queues (DLQs), and CloudWatch alarms to prevent data loss and provide real-time alerts.
</description><pubDate>Fri, 22 Nov 2024 00:00:00 GMT</pubDate><category>AWS</category><category>Lambda</category><category>Monitoring</category><category>SQS</category><category>DLQ</category></item><item><title>Secure Authentication and Authorization in SPAs: OpenID Connect and OAuth 2.0 with PKCE</title><link>https://bitvaria.com/en/secure-auth-spa-oauth/</link><guid isPermaLink="true">https://bitvaria.com/en/secure-auth-spa-oauth/</guid><description>A detailed guide to secure authentication and authorization strategies for Single-Page Applications (SPAs). OAuth 2.0 and OpenID Connect are combined with PKCE to ensure secure and efficient access control.</description><pubDate>Sat, 19 Oct 2024 00:00:00 GMT</pubDate><category>OAuth 2.0</category><category>PKCE</category><category>Authentication</category><category>Security Architecture</category></item><item><title>Beating Your Competition with Serverless: Accelerate Time-to-Market and Minimize Opportunity Costs</title><link>https://bitvaria.com/en/saving-money-with-serverless/</link><guid isPermaLink="true">https://bitvaria.com/en/saving-money-with-serverless/</guid><description>Discover how serverless computing helps you outpace competitors by speeding up development and avoiding opportunity costs.</description><pubDate>Tue, 15 Oct 2024 00:00:00 GMT</pubDate><category>serverless</category><category>cloud</category><category>web-applications</category><category>cost-optimization</category></item></channel></rss>