Interesting, I had always wanted a harness capable of model routing depending on the task, there are a lot of cases to keep in mind when bringing in model fusioning as mentioned in the last para s. Let's see how efficient and capable this is gonna be
Introducing Fusion in Devin CLI
The most efficient frontier harness for Fable & Astra; 39% cheaper across coding benchmarks.
Pick your favorite model for planning and a cost-effective model for execution.
Introducing Fusion in Devin CLI
The most efficient frontier harness for Fable & Astra; 39% cheaper across coding benchmarks.
Pick your favorite model for planning and a cost-effective model for execution.
Recently, I started learning about systems engineering, man it is so interesting, mostly on UEFI firmware and cpu, gpu and npu inferencing.
If u guys have any good blogs, playlist, channels for learning related to these, share pls
I’ve long maintained (for over 10 years atp) that the Indian education system is a scam. Honestly I’m tired of saying it again and again. Everyone knows the problem, but these days I have a different opinion on how to change it. Given the climate, it’s not a popular opinion but I’ll say it anyway. There’s actually two ways to do it and the best analogy I’ve found to explain it is actually from the anime Code Geass.
The first is to directly take on the system, through anger and being vocal about it. It’s the Lelouch approach. But over time you will find that in order to fight villains, you have to become a villain yourself. This path starts peacefully with a genuinely good objective but is often hijacked by people who just enjoy violence and don’t care about the mission. It’s still a necessary path and it’s important some people take it. And we have a constitutional right to raise our voices. Especially because students have lost their lives because of this system, it’s critical to be vocal and put pressure on it. Many of our own channels have been as vocal as we can. But my experience of seeing these over decades means that in a year or two we’re back in the same situation. Temporary solutions, temporary apologies. But a permanent problem with new faces.
There’s another path, the Suzaku path. The Suzaku path fully understands that the system is broken and there are villains. But the approach is to patiently and slowly change the system. In the real world that also means building viable alternatives. It’s the less flashy, less vocal but more effective path. Unlike the other one, it takes a long time, sometimes decades of doing things like creating jobs and slowly but surely influencing the system to produce the type of talent you need and thus also changing things in the process. The education system exists to serve the ecosystem of jobs so over time the system has no choice but to change.
Separately, I also think we really need better technology in our systems to prevent things like leaks. If people opt for the default path, we have a right to demand that it is at least fair. And a resignation to keep the system accountable is a correct ask. But at the same time we need alternatives for those who don’t make it in the current system. It’s brutal and unfair and I think many people are super talented but aren’t very good at giving an exam. We cannot judge a person based on just that. I hope to dedicate my life to helping build some of those alternatives and who knows, many one day that even becomes the norm.
Introducing https://t.co/RypxI9RDSH, expense splitting has to be this easy and seamless. These are the primary features,
- Temporary rooms (No auth)
- Permanent rooms (Oauth) - rolling out soon!
- Bill scanning
- Itemwise splitting
- Pay with UPI
- PWA
#BuildInPublic
Did a closed circle roll out, they turned out to be good beta testers lol, got a lot of feedback and improvements, constantly iterating to provide the best user experience.
Next focus, implement auth for persistent rooms
Building a startup taught me more about engineering
than any job ever did.
A job teaches you how to write good code.
A startup teaches you why the code even matters.
In a job,
you work on tickets.
In a startup,
you work on survival.
If your code breaks, users feel it.
If your system is slow, business suffers.
If you overengineer, money burns.
There’s no manager to hide behind.
No PRD to blindly follow.
No one to say, “not my problem”.
Everything is your problem.
Startup engineering forces you to think end to end.
Performance.
Cost.
Scalability.
UX.
Edge cases.
Failures at midnight. (sometime even on holidays)
You stop writing code just to look smart.
You start writing code that works.
You learn to make trade-offs.
You learn to ship imperfect things fast.
You learn to fix things while users are already using them.
And that pressure…
that responsibility…
It changes you.
It makes you calm.
It makes you practical.
It makes you sharp.
That’s why I say this openly:
Great engineers are not created by any tutorials.
They’re created by ownership.
If you really want to grow as an engineer,
build something that people actually use.
It will humble you.
It will scare you.
It will teach you more than any course or job ever will.
And once you experience that,
you’ll never look at engineering the same way again.
Save this if you’re building.
Share it with someone who wants to.
Cheers,
Akshay Saini
#StartupEngineering