<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>awaited.dev · Appunti pratici su IA e sviluppo software</title><description>Release di modelli, benchmark, esperimenti e guide per sviluppatori che lavorano con l’IA ogni giorno. Scritto a partire dall’uso reale, con fonti per ogni affermazione.</description><link>https://awaited.dev/it/</link><language>it</language><atom:link href="https://awaited.dev/it/rss.xml" rel="self" type="application/rss+xml"/><lastBuildDate>Sat, 01 Aug 2026 00:00:00 GMT</lastBuildDate><dc:creator>A.M.</dc:creator><generator>Astro</generator><docs>https://www.rssboard.org/rss-specification</docs><item><title>Ogni sessione parte da zero. Il mio agente riceve una mappa</title><link>https://awaited.dev/it/guides/context-management-for-coding-agents/</link><guid isPermaLink="true">https://awaited.dev/it/guides/context-management-for-coding-agents/</guid><description>La mia AI inizia ogni sessione senza memoria del repo. Ecco la mappa di file che ha risolto il problema: quattro livelli, un ordine di autorità, costi misurati.</description><pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/guides/context-management-for-coding-agents.png" medium="image" type="image/png" width="1200" height="630"/><category>guides</category><category>context-management</category><category>coding-agents</category><category>claude-code</category><category>agents-md</category><category>documentation</category><category>vibe-coding</category></item><item><title>Opus 5 pareggia con Sol, Terra eguaglia Fable a un quarto</title><link>https://awaited.dev/it/benchmarks/gpt-5-6-vs-claude-5-benchmarks/</link><guid isPermaLink="true">https://awaited.dev/it/benchmarks/gpt-5-6-vs-claude-5-benchmarks/</guid><description>Confronto Sol, Terra, Opus 5 e Fable 5 su benchmark di coding condivisi. I primi sono alla pari, Terra costa meno e più effort può ridurre la qualità.</description><pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/benchmarks/gpt-5-6-vs-claude-5-benchmarks.png" medium="image" type="image/png" width="1200" height="630"/><category>benchmarks</category><category>gpt-5-6</category><category>opus-5</category><category>fable-5</category><category>benchmarks</category><category>coding-agents</category><category>reasoning-effort</category></item><item><title>Kimi K3 eguaglia GPT-5.6, ma impiega 4,6 volte il tempo</title><link>https://awaited.dev/it/benchmarks/kimi-k3-glm-5-2-hype-check/</link><guid isPermaLink="true">https://awaited.dev/it/benchmarks/kimi-k3-glm-5-2-hype-check/</guid><description>Kimi K3 eguaglia GPT-5.6 nel Coding Agent Index, ma impiega 4,6 volte il tempo. I token economici di GLM-5.2 costano 6,51 $ per task tentato.</description><pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/benchmarks/kimi-k3-glm-5-2-hype-check.png" medium="image" type="image/png" width="1200" height="630"/><category>benchmarks</category><category>kimi-k3</category><category>glm-5-2</category><category>benchmarks</category><category>coding-agents</category><category>open-weights</category><category>pricing</category></item><item><title>GPT-5.6 Sol trova tutto. Ed è proprio questo il problema</title><link>https://awaited.dev/it/experiments/gpt-5-6-sol-overengineering/</link><guid isPermaLink="true">https://awaited.dev/it/experiments/gpt-5-6-sol-overengineering/</guid><description>GPT-5.6 Sol ha trovato 400 possibili problemi contro i 70 di Fable 5, ma la maggior parte non ha superato il triage. Spiego come limitarne il lavoro.</description><pubDate>Wed, 29 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/experiments/gpt-5-6-sol-overengineering.png" medium="image" type="image/png" width="1200" height="630"/><category>experiments</category><category>gpt-5-6</category><category>sol</category><category>openai</category><category>codex</category><category>fable-5</category><category>multi-agent</category></item><item><title>Convex a confronto con otto alternative. Convex resta</title><link>https://awaited.dev/it/guides/convex-alternatives/</link><guid isPermaLink="true">https://awaited.dev/it/guides/convex-alternatives/</guid><description>Gran parte dei miei progetti con l&apos;AI usa Convex. L&apos;ho confrontato con otto alternative, ho verificato i prezzi e trovato un costo nascosto per l&apos;UE.</description><pubDate>Wed, 29 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/guides/convex-alternatives.png" medium="image" type="image/png" width="1200" height="630"/><category>guides</category><category>convex</category><category>supabase</category><category>appwrite</category><category>insforge</category><category>backend</category><category>vibe-coding</category></item><item><title>Opus 5: ottimo dipendente, capo frustrante</title><link>https://awaited.dev/it/releases/opus-5-great-employee-bad-boss/</link><guid isPermaLink="true">https://awaited.dev/it/releases/opus-5-great-employee-bad-boss/</guid><description>Opus 5 eguaglia Fable 5 nei benchmark a metà prezzo per token, ma le lamentele della prima settimana sono reali. Cosa fa bene e dove gli serve un capo.</description><pubDate>Wed, 29 Jul 2026 00:00:00 GMT</pubDate><dc:creator>A.M.</dc:creator><media:content url="https://awaited.dev/og/it/releases/opus-5-great-employee-bad-boss.png" medium="image" type="image/png" width="1200" height="630"/><category>releases</category><category>opus-5</category><category>fable-5</category><category>anthropic</category><category>benchmarks</category><category>claude-code</category></item></channel></rss>