👨‍💻🏔️🏍️ Bringing The Future To Today | Founder | Software Engineer | CEO @ @echowinai, Building AGI @ @turnpanel Sharing what's on my mind.

Austin, TX
It’s great that we got the folding phone but when are we getting folding laptops?
26
5G these days is insane!
1
2
108
Kaushal Subedi 🧊 retweeted
I’m going to SF Tech Week. My agent on @turnpanel found the right events, booked them, built me an app to track it all, and made this video.
3
7
226
Congrats Google on finally shipping another frontier model, but I think shipping this to a limited amount of people when other labs made their models available to everyone is a big miss.
Introducing Gemini 4 Argon, our new frontier model, rolling out to cyber defenders starting today, and more widely as soon as possible. I am really excited by the progress we have made here. Argon is priced at $2 in and $10 out during introductory pricing!
1
89
Left @TMobile after 18 years on the network. Vote with your wallet folks 💳
67
SF is this in hard mode.
austin wealth is fascinating because the guy sitting next to you in shorts and a hat at matt’s el rancho could either be unemployed or worth $400 million there is absolutely no way to know
121
Kaushal Subedi 🧊 retweeted
We had 'Spaces' before ChatGPT. Our spaces live locally on your machine, not in the cloud. You can keep a space isolated, open it up, or collaborate with your team in it. Each space has its own jobs, skills, scripts, and custom apps you can build.
4
5
175
New version just dropped, make sure to update in time.
holy shit america 2 just dropped
1
124
Okay what if we go from names to usernames. Global identifiers for humans and agents.
1
72
Context
🚨 NOW: The Trump administration is launching a Super Intelligence site that will let Americans change their last name after marriage in ONE SINGLE FORM — which usually requires HOURS or DAYS to fix every government document "You'll be able to change your name right on the platform." "This would've taken weeks, multiple filings, different agency websites, if you even knew where to start." This is perfect! Saving Americans' time is INVALUABLE
27
Anthropic could do the funniest thing right now by increasing their usage limits.
1
47
If you didn't see this coming, you might need a new pair of glasses. 👓
Hi, Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan. Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago. (a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want. (b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions. (c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent. (d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet. I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news. Codexingly, Tibo
1
122
Kaushal Subedi 🧊 retweeted
"Slop" has become a comfort blanket for a cohort of programmers stuck somewhere between anger and bargaining on their way to acceptance. There might be some kicking and screaming, maybe a little crying, but eventually you have to let the blanket go and update your priors.
268
293
4,175
272,301
Kaushal Subedi 🧊 retweeted
TurnPanel now runs on Linux! 🐧 Set it up with one command on a VM, server, or DGX Spark. Then open your workspace from any browser. TurnPanel is now available on Windows, macOS, and Linux.
3
5
297
Ok so making polished video is a solved problem. Does this mean scrappy screen recordings will be more special now?
1
2
189
Kaushal Subedi 🧊 retweeted
What’s “coming” to Copilot is already here with @turnpanel , and it’s helping thousands of users get work done today. One workspace to write code, develop products, manage projects, build apps and run business operations, including accounting and payroll workflows. You can even build a custom app to manage a workflow that doesn’t fit the tools you have. TurnPanel can use your browser, files, terminal, local apps, and online services, whatever is needed to complete a job. It can move across those resources instead of stopping at the edge of one app. You can access agents running on multiple computers from one interface, even on your phone when you’re headed to the beach. You can create spaces for different parts of your business, each with its own context, memory, skills, scripts, and workflows. Your agent evolves with you, and your work continues to get easier and more cohesive over time. More importantly, the workspace sits on your machine. Your files, memory, and workflows live there, and you own them. That matters. Pick an online model, run a local model if your machine supports it, bring your own model, or connect an OpenAI-compatible provider. All from the same workspace. Build the app. Ship the product. Run the operation. Do it from the same place, with the context and tools to finish the job.
We’re building Copilot as a new OS for work that spans every model, every form factor, and every task. Today, we’re announcing our biggest update to Copilot to date, bringing four things together: · Autopilot: proactive and long-running agent built for the enterprise · Code: build apps with Copilot, hosted inside your company’s tenant · Home: Chat + Cowork together · Office: now fully embedded in Copilot (and Copilot embedded in Office, of course!) Plus, you can invoke Copilot in Teams, and we’re introducing Today, a proactive experience that surfaces the most important information from across M365 without needing to ask for it. The way we work is changing and so are our workflows. This update brings AI into that flow, from answering a question, to building an app, to getting work done on your behalf.
3
2
5
831
Kaushal Subedi 🧊 retweeted
Bring your work (and intelligence) home with TurnPanel
3
8
263
Kaushal Subedi 🧊 retweeted
TurnPanel is live on Product Hunt! 🚀 We built it for people who want AI to do more than answer questions: help get real work done across their computer, with a local-first approach and the user in control. If that sounds useful, take a look at our launch and tell us what you think. Your feedback would mean a lot to our team.
3
5
10
290
While I love that there is an open source alternative to Jev, the folks saying that Laya is the same have no clue how different these models are. Here is a benchmark we ran comparing classification task on MARS (our in-house classification SLM with 250M parameters), Jev and Laya (out of the box). As you can see, Laya is not even close but Jev gets pretty close out of the box.
1
4
252
Then we thought, Laya is a small model, maybe we can fine tune it to make it match our SLM with performance benefits. We then trained Laya on a similar dataset to the one we used to train MARS. The results were much better, we got Laya very close to both MARS and Jev, but it was still not perfect.
1
1
2
108
But this is where things got interesting. When we compared the resource usage, the 250M SLM was not only smaller, but it was also faster and more accurate (by a tiny margin) than Laya. So it seems like training a small language model is still better for classification over something like Laya, at least for now. I am excited to see how these models evolve over time though.
1
1
49