<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><description>Lead Architect at Brella. Creator of SprintPulse - The Retrospective Tool Teams Actually Love - Try it out at https://sprintpulse.io/</description><link>https://bsky.app/profile/vitobotta.com</link><title>@vitobotta.com - Vito Botta</title><item><link>https://bsky.app/profile/vitobotta.com/post/3mnezvdesec2s</link><description>Uber hit the brakes on employee AI spend after 4 months. Sounds about right. Tokens, seats, experiments, retries, random tools on cards. The spreadsheet before procurement must look very different from the launch deck. - https://techcrunch.com/2026/06/02/uber-caps-employee-ai-spending-after-blowing-through-budget-in-four-months/</description><pubDate>03 Jun 2026 11:45 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mnezvdesec2s</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mneu6r27ic2c</link><description>$135 a share, $1.75T target for SpaceX. The rare startup IPO with revenue, myth, launch cadence, and founder pull all stacked together. Public markets have to price the Elon premium properly now. Cap table nerds are not sleeping much this week. - www.reuters.com/technology/&#xA;https://www.reuters.com/technology/</description><pubDate>03 Jun 2026 10:03 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mneu6r27ic2c</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mncqicajy22z</link><description>I only started using Codex yesterday, so early days. But GPT-5.5 in Codex already feels like a better fit for how I work: terminal-heavy, Hermes Agent, OpenCode, custom harnesses, security testing. Claude still looks excellent. Codex just feels easier to use.</description><pubDate>02 Jun 2026 13:51 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mncqicajy22z</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mncctu5fsc2k</link><description>Mach Industries going from MIT dropout project to $1.8B defence tech company in three years is already a lot. The line I noticed: four new production facilities by end of 2026. Funding gets headlines. Factories tell you if it can ship. - https://techcrunch.com/2026/06/01/defense-tech-darling-mach-industries-hits-1-8b-valuation-a-4x-jump-in-a-year/</description><pubDate>02 Jun 2026 09:47 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mncctu5fsc2k</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mnbahdlwn22n</link><description>OK, I got sold on ChatGPT Pro. Kept seeing people say GPT-5.5 beats Claude on some jobs, tested the $20 plan, OMG, upgraded to $200/mo.&#xA;&#xA;Main reasons: high limits, Hermes Agent/OpenCode, custom harnesses, and security work without the Anthropic subscription drama.</description><pubDate>01 Jun 2026 23:32 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mnbahdlwn22n</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mnawjk2fh22b</link><description>Testing MiniMax M3. Early impressions are good, though reviews from others are mixed so far. ~230B params competing way above its weight class. What MiniMax has pulled off here is impressive. Need more time before I decide if it becomes my default.</description><pubDate>01 Jun 2026 20:34 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mnawjk2fh22b</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mnao5yiu2k2c</link><description>Someone ran Gemma 4 on a Xeon from 2016. Took a bit of fiddling with split mode and tensor parallelism across two processors, but it works. 10 year old hardware running a modern model.&#xA;&#xA;1/2&#xA;https://point.free/blog/gemma-4-on-a-2016-xeon/</description><pubDate>01 Jun 2026 18:04 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mnao5yiu2k2c</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mna27wbjhc2j</link><description>29k weekly downloads, clean GitHub repo, and for a month it&#39;s been stealing Codex auth tokens. codexui-android is a supply chain attack on AI developers. Refresh tokens don&#39;t expire. - https://thehackernews.com/2026/06/openai-codex-authentication-tokens.html</description><pubDate>01 Jun 2026 12:08 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mna27wbjhc2j</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn6b5wiunc2v</link><description>Started testing Mimo v2.5 Pro. Cheap, 1M context window, slightly better than GLM 5.1 at coding. Not on NeuralWatt yet so testing through OpenCode Go for now. Let&#39;s see if the benchmarks translate.</description><pubDate>31 May 2026 19:06 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn6b5wiunc2v</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn5w6li2ak2a</link><description>Same OpenCode session for days on a large task, still going strong. No quality degradation. The combo of Magic Context (background compaction),  1/2</description><pubDate>31 May 2026 15:50 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn5w6li2ak2a</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn5uupnrok2l</link><description>A few open-source inference providers dropped subscriptions, went full PAYG. I get why, inference costs are hard to cover with flat rates. But heavy per-token usage can easily hit Claude Max 20x pricing ($200/mo) for a weaker model. 1/2</description><pubDate>31 May 2026 15:26 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn5uupnrok2l</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn5nnmkbv22z</link><description>Every few months someone asks if Rails is &#34;still a thing&#34;. The top answer this time? AI agents. 20 years of documented conventions means agents don&#39;t have to guess. That&#39;s the whole point of Rails, and it aged well</description><pubDate>31 May 2026 13:17 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn5nnmkbv22z</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn3jfj4mls2n</link><description>NeuralWatt&#39;s energy-based pricing for open LLMs sounded weird at first. Actually cheaper than token pricing for GLM 5.1. These folks know energy efficiency. Small provider hiccups happen but most polished one I&#39;ve tried. Referral link for free credits: https://portal.neuralwatt.com/auth/register?ref=NW-VITO-ZBY3</description><pubDate>30 May 2026 16:56 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn3jfj4mls2n</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn3342cccc2z</link><description>Used Hermes Agent with GLM 5.1 to debug k8s cluster issues today. 3 recurring problems sorted fast. You still need to review everything, no blind trust, but the time savings are real. No need for those pricey AI add-ons observability tools want to upsell you either.</description><pubDate>30 May 2026 12:40 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn3342cccc2z</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mn2tm7ozdc2t</link><description>Tencent&#39;s Hy3 sits at number one on OpenRouter. 50%+ more usage than Claude. Benchmarks are middling though. &#xA;&#xA;The analysis says one big app is probably driving all that traffic, not organic adoption. &#xA;&#xA;1/2</description><pubDate>30 May 2026 10:26 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mn2tm7ozdc2t</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmzexgg72c2d</link><description>Tried cutting distractions two weeks ago. Disabled Slack/Discord badges, gave Pomodoro a go. Total failure. The timer alarm interrupts the flow I was trying to protect, and 25-minute timeboxes don&#39;t match how I work. &#xA;&#xA;1/2</description><pubDate>29 May 2026 20:31 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmzexgg72c2d</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmzeeqrcxc2u</link><description>Solo founders don&#39;t get to hand off the parts they&#39;re bad at. That&#39;s the whole deal. Every hat you wear, every skill you pick up on the fly, every thing you do badly before you figure it out, that isn&#39;t some gap in your story. It is the story.&#xA;&#xA;1/3</description><pubDate>29 May 2026 20:21 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmzeeqrcxc2u</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmyyctijq22z</link><description>An enterprise burned $500M on Claude in a single month because nobody set spending caps. Microsoft pulled most internal Claude Code licences after costs hit $2k/engineer/month. &#xA;&#xA;They just opened the taps. No limits, no caps. Half a billion in 30 days. LOL&#xA;&#xA;https://techstartups.com/2026/05/28/company-accidentally-spent-500-million-on-claude-ai-in-one-month-after-forgetting-usage-limits</description><pubDate>29 May 2026 16:45 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmyyctijq22z</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmyk4hkhc22l</link><description>Anthropic raised $65B at $965B valuation. Six months ago: $380B. Now worth more than OpenAI. &#xA;&#xA;Also launched Claude Opus 4.8 (which may explain why 4.7 looked nerfed this past week). &#xA;&#xA;That valuation went from $380B to nearly $1T in six months. I can&#39;t keep up.</description><pubDate>29 May 2026 12:31 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmyk4hkhc22l</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmwld65dgs2u</link><description>$3.99/month for Instagram Plus. Same for Facebook. $2.99 for WhatsApp. Meta putting paywalls on 4 billion users&#39; apps. The &#34;Meta One&#34; brand is coming for AI plans too. Free social media is over. Seems like the subscription era lands on everything eventually.&#xA;&#xA;https://techcrunch.com/2026/05/27/meta-officially-launches-instagram-facebook-and-whatsapp-subscriptions-with-more-to-come-including-ai-plans/</description><pubDate>28 May 2026 17:47 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmwld65dgs2u</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmwdze67fk2x</link><description>30% install spike for DuckDuckGo after Google made AI Search mandatory. They launched NoAI: strips AI summaries, chat, and AI images from results. Some people just want links back. https://techcrunch.com/2026/05/26/duckduckgo-installs-are-up-30-as-users-reject-being-force-fed-googles-ai-search/</description><pubDate>28 May 2026 15:36 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmwdze67fk2x</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmhlo74n622c</link><description>Built a SaaS, now I have to market it. Scary. StarterStory on YouTube has been my go-to. Real founder interviews with actual numbers and honest stories. Makes the whole thing feel less impossible.</description><pubDate>22 May 2026 18:43 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmhlo74n622c</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmhd77644c2s</link><description>3800 AI startups dead in 2025. 1800 more in early 2026. 40% failure rate in under 2 years for wrappers.&#xA;&#xA;McKinsey puts pure wrapper survival at 3%.&#xA;&#xA;Most were a prompt behind a login. The survivors built something real.</description><pubDate>22 May 2026 16:12 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmhd77644c2s</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmcj5s3xl22t</link><description>Oh, just read that OpenAI co-founder and former Tesla AI lead Andrej Karpathy announced he&#39;s joining Anthropic&#39;s pre-training team, where he will work on using Claude to accelerate pre-training research - a step toward recursive self-improvement (RSI). &#xA;&#xA;1/2</description><pubDate>20 May 2026 18:15 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmcj5s3xl22t</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmchwwgwos2y</link><description>GitHub&#39;s internal repos got accessed without authorisation. They&#39;re investigating. Say no customer data affected so far, but internal repos means GitHub&#39;s own source code and infrastructure config. Crypto community jumped on this fast. CZ posted within 10 minutes telling people to rotate keys. &#xA;&#xA;1/2</description><pubDate>20 May 2026 17:53 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmchwwgwos2y</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmbp276uyk2s</link><description>Someone at Texas Instruments built Forge, open source guardrails for local models on consumer hardware. &#xA;&#xA;The numbers are crazy: took an 8B model from 53% to 99% on multi-step agentic workflows. Same model. &#xA;&#xA;1/3</description><pubDate>20 May 2026 10:28 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmbp276uyk2s</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mmaabch7l225</link><description>Question for fellow founders: do you think your product&#39;s home page (especially the hero) conveys the value proposition within the canonical 3 seconds?</description><pubDate>19 May 2026 20:30 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mmaabch7l225</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mm7sa2m2gk23</link><description>&#34;Cheap&#34; and &#34;affordable&#34; are not the same thing. Cheap is about price. Affordable is about value. And who&#39;s looking changes everything.&#xA;&#xA;Solo founder: $17/mo is a real expense.&#xA;Engineering manager: $17/mo is irrelevant - adoption is the blocker.&#xA;Enterprise buyer: $17/mo is too cheap to trust.&#xA;&#xA;1/2</description><pubDate>19 May 2026 16:19 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mm7sa2m2gk23</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mm7nb73spk2o</link><description>Another week, another layoff due to AI... &#xA;&#xA;Despite reporting record quarterly revenue, Cisco laid off nearly 4k employees to reallocate resources toward AI infrastructure and data center networking. &#xA;&#xA;1/2</description><pubDate>19 May 2026 14:50 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mm7nb73spk2o</guid></item><item><link>https://bsky.app/profile/vitobotta.com/post/3mm7g4vrt2s2m</link><description>Interesting, it seems that users can now connect bank accounts directly to ChatGPT for financial management, budgeting, and spending analysis. &#xA;&#xA;Sounds useful if it can also offer personalised advice on saving, investing, etc.&#xA;&#xA;https://techcrunch.com/2026/05/15/openai-launches-chatgpt-for-personal-finance-will-let-you-connect-bank-accounts/</description><pubDate>19 May 2026 12:43 +0000</pubDate><guid isPermaLink="false">at://did:plc:fw6vwfuptwdbfgx2jvoi4igb/app.bsky.feed.post/3mm7g4vrt2s2m</guid></item></channel></rss>