Building Intelligence - ai/acc (@onlygpus) header
Building Intelligence - ai/acc (@onlygpus) profile picture

Contributor to Spheron

Remote
Joined October 2025
Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)Photo by Building Intelligence - ai/acc (@onlygpus)
I am building a local whisper as a side project for my laptop to enhance security and customization. Here are the reasons behind this decision: - Trusting a third-party app with sensitive information is risky, especially given recent security incidents around the globe. - By installing locally, I can customize the software to better suit my specific needs, rather than relying on what is offered by third parties. - My keystrokes are kept private when using a local model on my Mac, which now supports dictation effectively. - When using external models, I anonymize everything confidential and only send non-sensitive data to external APIs. If you want to try it out on your local machine, feel free to send me a DM, and I will share more details with you. - above written by local model & dictation
3
93
We’ve secured 200 B300 nodes for February delivery and another 40 racks of GB300 going live in March, all in Tier 3+ data centers in APAC. We can offer a 1-year commitment with a 99.8% SLA. If you’re planning GPU capacity for 2027, let’s talk.
3
1
2
254
Building Intelligence - ai/acc retweeted
Your AI agent can now rent its own GPU. GPU Sandbox is live on Spheron. Agents rent GPUs, train, fine-tune, run inference and manage production workloads, without you handing over SSH keys. Works with @claudeai or any agent harness.
2
1
5
1,564
we have got 2 nodes of B300 available in thailand, if anyone is interested we are offering 3 months commit of these machines. my DM's are open
3
1
141
We use AI in every part of Spheron: code, code review, content, posting, signal-hunting on socials, emails, onboarding and daily customer updates. Video was the last gap. After months of testing, we cracked it. Meet our new AI avatar hire 😎 She’s the first of a few presenters for our explainer videos. Not low-effort AI content, but real explainers worth your time. Episode 1: why @spheronai is all boxes and blue. Made end to end with AI. If you like it, show some love 🙌
1
1
75
For smaller teams, owning GPUs rarely makes sense. What you’re building today could look completely different a year from now and so could your compute needs. Rent the capacity you need today. Keep the flexibility to change tomorrow.
1
32
Building Intelligence - ai/acc retweeted
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 github.com/deepseek-ai/DeepI… 1/4
1,766
4,939
2,351
43,870
7,112,386
Still waiting for GPU compute to get cheaper? The same people telling you to wait are busy locking in capacity for themselves. You’re being sold a story while they’re buying the supply. 😅 By January, you could be wishing you’d locked in today’s rates.
you should scramble to get compute because I have a feeling it's about to get ugly
59
Building Intelligence - ai/acc retweeted
B200 spot runs $0.073 per million tokens. H100 PCIe on-demand runs $0.451. Same workload, same checkpoint, a 6.2x spread. MLPerf Inference v6.0 (April 2026): B200 hits about 17,500 tok/s on Llama 2 70B against H200's 7,800. On GPT-OSS 120B, a 117B-param MoE with only 5.1B active per token, that gap widens to 2.5x. MI355X carries more memory than both, 288GB HBM3e vs B200's 192GB, and still trails at roughly 2,600 tok/s on the same GPT-OSS test. Memory capacity and inference speed are not the same leaderboard.
2
1
3
797
Most of the GPU capacity the world needs is already built. It just isn’t connected. Look at where compute actually sits today: → A data center in Frankfurt with H200s and no sales team. → A cloud region in the US with H100s and a waitlist. → A colo in Oslo with B200s coming online and nobody who knows. → A lab with A100s idle 60 percent of the week. → A reseller holding “stock” that may or may not exist. Each one is an island. Each one is sold separately, through its own brokers, its own paperwork, its own price nobody else can see. An AI team that needs 512 GPUs for 12 months doesn’t want five negotiations. They want one answer. That’s the whole idea behind Spheron: one grid, one gateway, one marketplace. Every data center, cloud and cluster becomes a node. One marketplace to find it, one API to run on it, one MNDA to reserve it, one bill at the end. The supplier keeps the customer. We keep the grid running. Electricity went through this a century ago. Thousands of isolated generators, then a grid, then nobody thinks about where the power comes from. Compute is at the generator stage. We’re building the grid. Where is your compute sitting right now, and how many separate contracts does it take to use it?
1
1
76
Building Intelligence - ai/acc retweeted
Don’t buy GPUs, buy access. Hardware ages. Access scales. With @spheron, you get on-demand enterprise-grade compute that grows with your workload, no sunk costs, no hardware hassle. $SPON gives you scalable AI power, without ever owning a single GPU.
4
2
9
3,332
Building Intelligence - ai/acc retweeted
No guess work. No emotions. Just data. Myra is changing the way you trade - powered by @Aster_DEX , live on @BNBCHAIN . Build your circle, share strategies, and let your community win with you. Sign up today and trade smarter, together.
6
5
1
14
79,860
Building Intelligence - ai/acc retweeted
Just bought some GPU power from @spheron — as our GPU demand keeps rising, we ran multiple comparisons with other providers and found that Spheron offers much cheaper and more reliable GPU power.
5
4
2
24
7,289
DEX + Perp Volume + Active Addresses (24h) Solana: $5.7B, 2.53M users BSC: $4.5B, 2.85M users Base: $1.8B, 953K users Polygon: $520M, 517K users Solana leads trading, BSC owns user count. Base grows, Polygon waits for its moment.
1
51
Takeaway: @solana app ecosystem is pulling ahead, @BNBCHAIN drives retail, @base builds momentum, and @0xPolygon lays groundwork. The real story? A multi-chain world is forming-where seamless cross-chain UX
40
Building Intelligence - ai/acc retweeted
DIN, the Decentralized Infrastructure network, is poised to become always-on dialtone for the decentralized economy. Born of and continuously nurtured by @infura_io, @DINBuild is maturing rapidly, has accumulated many frens and partners and is progressively decentralizing. Oh, and @MetaMask already uses it extensively. So yes, institutional grade.
Replying to @ethereumJoseph
we need solid dRPC network that works 😅
29
17
6
162
34,958
Building Intelligence - ai/acc retweeted
like if you've built something onchain. RT if you're building something this week. reply with what you're working on.
1,067
805
75
3,676
229,207