Digital products since 2006. Invited by ClickBank to speak. Building https://xitter.cf/t.co/OhiZSN36Iy: journal your life → ideas you haven't had yet. Gacha is my one weakness.
Bangkok, Thailand
Joined May 2008
- Tweets1.2K
- Following325
- Followers286
- Likes799
If you are looking for a great local LLM packaged in a box (LLM + optimization layers + dashboard + chat, ...)
I think @0xBakeer's TandemLLM is one of the best packaging right now.
Don't just see it, try it.
Base on the history of Apple launch events, when Apple want to release a brand new hardware, they will secretly work with a few major 3rd-party developers to show the capabilities of the new devices in the event.
Such as when they launch iPhone 3G with accelerometer showing off tilt-control, they have Super Monkey Ball developed by SEGA and showing to the world in the launch event.
Apple always work with developers closely before launching devices to make the best user experiences when it launches.
But now, Apple seems really don't care about this and follow the cheap tricks, release first, fix later...
I think a lot of people will miss the legendary Steve Jobs so much.
And this is a HUGE gap for indie hackers.
Get your feet wet with @bitrig now!
WhatToVibe.com is really riding on the trend and it proposes me a quite interesting idea to build!🤣
You got one? @ifourth
I just began developing my own task-specific local inference engine for 4-5 days, then DeepSeek released DeepInfer while NVIDIA just released Flare.
wow, so many things to learn from theirs.
NVIDIA just released Flare, a local inference server built for DGX Spark and RTX Blackwell GPUs
- Built for long chats, agents and tool calls
- Serves from Rust, no Python
- Plans memory up front
- FP8 KV cache by default
Interesting timing... right around when DeepSeek open-sourced their inference engine.
Link below.
Our plan-A feeding WhatToVibe.com to AI starting to works! @ifourth
ChatGPT is the first one 🔥🔥
Organic traffic is always welcome!
Don't stop building... quality will always found you a quality one. 🔥🔥🔥
Running new FP4 QAD quant of Qwen3.8-Flash-Next with TensorFold now.
Let's see how does it performs by throwing them a real work for 24 hours non-stop!
Rejoice single DGX Spark owners!
You can now run Qwen3.8 Flash Next on TensorFold with the new Nvidia FP4 QAD quant 🔥
- 1.1M total FP8 KV cache
- 256k context, 5 concurrent users
- 46-55 tok/s on prose, single stream
- 132 tok/s on prose, 5 concurrent steams
- Prefill 2000-2300 tok/s
Qwen3.8 Flash Next is still the BEST model to run on one DGX Spark, bare none!
You can choose what to run: FP4 QAD for quality, or int4 for speed. Details on what is QAD below.
Get it here:
mia-ai.net/models/Qwen3.8-Fl…
First day move from TelemetryDeck to Amplitude.
It's time to map the best user journey and locate the worst bottleneck!
We got 6 ideas from WhatToVibe.com to build apps riding on Tesla flying roadster trend.
I like one of them, will tweak the idea a bit and send to my agent to build it.
Can’t be patient for T3 Code nightly anymore, running on Omarchy and it crashed 1-3 times a day randomly. 😑
Returned to Orca even I prefer T3 phone app more.
I reverse engineered reloop.sh, and it's 49% cheaper.
✨📷 Introducing relube.sh 📷📷
Everything you need to send email from your app...
📷 Email API
📷 Inboxes for your AI agents
📷 Automations
📷 Deliverability guardrails
...
Nooooo, it's just a joke. 😂😂
I reverse engineered Resend, and it's 78% cheaper.
✨✨ Introducing reloop.sh ✨✨
Everything you need to send email from your app...
⚡ Email API
📮 SMTP relay
📨 Inboxes for your AI agents
🔁 Automations
🛡️ Deliverability guardrails
🔗 Bring your own SMTP (and keep your reputation)
And it's not a dinosaur tool:
+ It's open source; you can run it for free
+ It's agentic; everything in the UI is available to your AI
Move over from Resend in 2 minutes.
100k emails for $20/month, not $90. Free up to 3,000/month.
Oh, and it receives email too, not just sends it.
Never thought that @0xBakeer's TandemLLM is super easy to setup.
Very relief to skip terminal for a while. 😆
Will battle test with my real works after it lives!
BTW: I really love his concept of TandemLLM, especially the layers that work together with open-weight models to produce a better control/workflow.
My favorite frontier model right now -> Claude Haiku 5.5
- Very fast
- Very good quality
- Virtually unlimited on my $100 max plan
