Nearly three decades across enterprise technology, finance, government, telecommunications, cloud, and executive leadership.

Replying to @sama @tomieinlove
That isn’t the cost of being polite, neither to humans nor to AI. We teach children to say “thank you”, “please”, not because it’s efficient, but because it instills empathy, respect, and emotional awareness. Technically, skipping polite phrases every day could speed up lines, shorten conversations, and even save time and money at scale. But we still say them. Why? Because human communication isn’t just about utility, it’s about connection. As AI becomes part of everyday interaction, how we speak (prompt) to it reflects our values. If we normalize blunt, transactional language for the sake of efficiency, we risk modeling a colder, less empathetic society. Courtesy isn’t a waste, it’s an investment in the kind of future we want, both with each other and with the intelligence we’re creating.
4
1
36
3,984
Arman Obosyan retweeted
Armenia can build AI companies with global reach. Computing infrastructure, talent, and entrepreneurship need to grow together. I’ll join Friends of Yerevan State University in Burbank on October 7 to discuss what it takes. 🇦🇲 Event details: linkedin.com/feed/update/urn…
5
10
84
2,481
Back for another year at Microsoft Global Hackathon 2026, this time with Live Grounding Layer for Continuous Agent Improvement, connecting AI agents to live, verifiable data through #SugraAPI Less guessing. More grounding. Thank you, The Microsoft Garage! @msftgarage credly.com/badges/5dae7329-1…
1
119
#ClaudeCode interrupted my coding today to invite me to an interview about AI. Interesting questions, and apparently there’s a portal where you can join directly @AnthropicAI Usually, I’m the one asking AI questions. Today, it was asking me. 🙂
1
3
158
Arman Obosyan retweeted
Introducing Gemini 4 Argon – our new frontier model. It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
1,968
4,636
42,912
8,750,567
Arman Obosyan retweeted
DataFest Tbilisi is more than talks & panels. 🎤 Talks & Panels 🛠 Workshops & Labs ⚡ Datathons & Hackathons 🤝 Networking & Side Events Learn. Experiment. Collaborate. Connect. More than one way to experience #DataFestTbilisi2026
1
1
67
Arman Obosyan retweeted
Покидайте ссылки на свои Saas. У всех же есть как минимум один в этоху ai? 🧐
29
2
37
9,951
Woke up to 62,500 credits from OpenAI. That’s $2,500 worth of fuel for Work and Codex. Apparently, the reward for using too much AI is more AI. I’m happy with this arrangement 😅 Thanks @OpenAI! Let’s build something cool #Codex
1
129
GPT‑6.1‑Sol Ultra is now in my #Codex session. Let’s see what we can make. @OpenAI Maybe the new Renaissance starts with four words: “Let’s build something cool”
1
1
2
127
7.4 billion tokens with Codex. Peak day: 489.2 million. Current streak: 41 days. At this point, I’m not sure whether I’m using Codex or stress-testing it. Apparently, I’ve been benchmarking OpenAI’s infrastructure. @OpenAI Either way, one thing is clear: It’s part of the engineering stack. ...and that’s just #Codex There’s a lot more happening in Claude and xAI 🤭
3
103
More creators. Fewer cogs. “I believe that the future, if we get this right, can be much more like a new Renaissance than a new Industrial Revolution. @sama at @OpenAI The most important words? “A new Renaissance” That’s the ambition we need #OpenAIDevDay
1
254
Useful reading, on AI tokens cache reads. We had our own numbers on 108.8B of 112.4B Claude Code tokens were cache reads, almost 97% url.sugra.blog/2of6
My AI agents wrote 0.4 million tokens of output last week. I paid for 70 million. That's not a billing error. It's how agent work is billed, and it changes which model is really "cheap". An open model I picked to save money turned out to cost as much as Claude Opus, just because of where it was hosted. Field notes from a week of letting Kimi, DeepSeek, GLM and Qwen work on my real app: escamoteur.dev/blog/how-good…
1
2
193
2 companies, 1 CEO. 20 #GrokBot’s, Each @bot with one job: dev, ops, research, regulation, mail, calls. I just review in the morning. Watching @jetsemrick's "Give each Bot a job" in Tbilisi felt familiar 😄 @nur_daulet_ @IgorBlinkk #SpaceXAI
2
228
"Give each Bot a job" 🤖 @jetsemrick at the Grok Bot Meetup in Tbilisi 🇬🇪 GTM, engineering, marketing, admin: bots do the busywork overnight, humans review in the morning 👀 @bot #GrokBot #Tbilisi #SpaceXAI
1
1
5
223
Grok Bot just landed in Tbilisi 🇬🇪 First meetup at @BTUuniversity, full room, no idea yet what we'll leave with. That's the fun part 🚀 @nur_daulet_ @bot #GrokBot #Tbilisi #SpaceXAI
1
1
3
157
👏🏼👌🏼🚀
The Sugra API MCP server is now listed in Anthropic's Connectors Directory url.sugra.ai/claude One connector, 1,600+ endpoints across 36 domains, in Claude'ai, Desktop, Mobile, Claude Code #SugraAPI #MCP #DevTools
73
Burn the entire weekly #SuperGrok Heavy limit? Easy! Done! and 96% of it on #GrokBuild. At this point,@grok, I think I deserve either a badge or a larger limit 🤭 #xAI #AICoding
1
1
96