Armagan Amcalar

It’s been 15 years since I last wrote an academic paper and I didn't expect the next to be this fundamental. Today we published a research paper with @openservai that represents a major breakthrough in how we think about AI reasoning. For years, we’ve been told that better AI reasoning means bigger models, longer chains of thought, and higher costs. We accepted that trade-off almost by default. This work questions that assumption. With BRAID's structured, bounded reasoning, we’re seeing up to 74× efficiency gains and ~30× better performance per dollar without losing accuracy.

Şu an Armağan Amcalar ile system prompt’ların executable logic’e dönüşümünü konuşuyoruz. 🧠 #diva #kadınyazılımcı #ai #yapayzeka @dashersw

Diva: Dive into AI Conference is live today.

What happens when you post a real Monet and say it’s AI? The coolest art social experiment I’ve seen in a while. Thank you @SHL0MS

@Cryptopuss @openservai Thank you!

@dashersw Proud to be a Investor in $serv since the beginning of it all. I will always support this team, it has been a incredible journey so far! So much more to come, let's get it! Thank you Armagan💪🏻 @openservai

This. I'm so proud of my team pulling this through!

When building and launching their project, @momus_ai , @snipe1prins and @kllrbeez could have chosen many teams to work with, but they chose @openservai! $Momus launches May 4th on the $SERV Launchpad. Watch this short clip from the #PTPodcast to see why they chose @open_founder, @lucashafner, and @dashersw make their idea come to life!

SERV Reasoning makes DeepSeek-v4-flash deliver 36× performance-per-dollar of GPT-5.5. • GPT-5.5: 79.46 score = $1.69 • GPT-5.4: 82.03 score = $1.05 • DeepSeek-v4-flash + SERV: 90.78 score = $0.05 Higher score, at a fraction of the cost. Private Beta opens soon. Apply below. 👇

@DanShappir @solid_js @sveltejs Thank you Dan! From the ideology perspective, it’s “compiled” like Svelte, but yes, it uses JSX for the templates. Where it fundamentally differs from Solid is that there is no special syntax or new concepts for state management.

@dashersw Cool! From a quick look it seems more like @solid_js than @sveltejs, e.g. tsx. Or am I missing something?

I finally released Gea v1.3.0 after 3 weeks of work. Now it's fulfilling the original promise of Svelte, in a way Svelte never did. The 172 byes hello world in Gea is 635 bytes in Svelte 1.0 and 8,537 bytes in Svelte 5.55.5. Gea is truly the magical disappearing framework. The reality is, we can keep the sane development practices based on JSX everyone is by now used to, and add ZERO mental overhead on top, and still compile away ALL of the parts you didn't need in the runtime. Give Gea a spin. It's the fastest framework on the market with the smallest bundles, that doesn't ask you to write weird, cryptic code. It's Just JavaScript™.

And first paint is even faster than vanilla.

Gea tops the charts in memory usage as well.

Burcu Yarar “Kurallar Değişti: Yapay Zeka Sistemlerini Hacklemezsiniz, İkna Edersiniz” konuşması ile bizimle olacak! 🎉 ⚠️ Saldırganlar sistemleri hacklemekten çok onları yönlendirmeyi öğreniyor. 🧠 Bir sistem yanlış kararlar almaya ikna edilebiliyorsa gerçekten güvenli midir?

İki Manifestonun Hikayesi: Palantir ve OpenAI nasıl bir dünya hayal ediyor? | 24 Nisan Cuma 21.00 x.com/i/broadcasts/1…

Sneak peek from our internal benchmarking tool. We're past 100 iterations on the SERV Reasoning engine. The engine gets better every cycle. Latest run, 18 DeFi scenarios, 100 runs: > gpt-5.4-nano with SERV Reasoning scored higher than vanilla 5.4 > ~10x cheaper > same model, same decoding, just better reasoning underneath SERV Reasoning performs better than frontier models, at a fraction of the cost. If you're building agents that need to make real decisions under real constraints, this is the layer you want underneath them. Early access for selected teams. Spots are limited. Sign up below. 👇 openserv.typeform.com/to/dG8koJgd

Cursor is the best coding harness ever built. Change my mind.

Claude is just full of hallucinations...

The $SERV livestream last week was filled with alpha (Reasoning execution sharding!?, Shadow Agents!?, How existing LLMs are broken!?) Don't have time to listen? I did a summary of the livestream. TLDR thread, and also why @openservai is a +250x research opportunity. 🧵👇👀

years ago @OpenAI started as a simple AI research lab, small team, @elonmusk and @sama, no products, just raw research for the benefit of humanity today @openservai is building something similar, they’re a small crypto-based AI research lab that is actually shipping killer products right now, plus the $SERV token currently sits at an FDV just under $20M tbh, it’s too late to invest in @OpenAI and @AnthropicAI, but with $SERV, you’re still incredibly early the first @openservai breakthrough was SERV Reasoning, a structured approach that laser-focuses outputs from frontier models, making API calls 74x cheaper and 3x faster the tech is already powering infrastructure for the @UAE, along with dozens of other governments and elite corporations, many still hidden behind NDAs the second invention is Build-Launch-Run, got an idea, agents help you write the code, the @openservai launchpad deploys it on @base, and soon @solana, then a full team of AI agents help you run the project via chat in a clean @SlackHQ style interface tokenomics are also best in class, 25% of all revenue goes toward buybacks, hold 50k $SERV and gain early access to every token launched on the platform, 5% of every token’s supply flows back to stakers staking is also expected to go live very soon, but obv buying $SERV right now on either @ethereum or @base allows you to lock in at the current valuation longterm they’re building an L3 chain and an LLM they believe could overtake the current field, small lab, massive ambitions, tiny FDV the labs that research the hardest tend to outperform, bc they gain access to the freshest innovations first, imho if feels like @openservai is establishing itself as the leader among next-gen AI research labs think of this, if the valuation of $SERV ever caught up to @openal, that would equate to a gain of more than 50,000x srsly, the upside is massive 👀🔥

I'm just sick of @claudeai for randomly stopping working. This is a huge blocker for progress. Why is it slacking off this much? Like, dude, your job is to code.