nathanstweeting

Some Software Development, some DevOps. Building things on the side, hope to gain financial independence. If you don’t enjoy the process, what’s the point?

[nathanstweeting @nnolantweets]

Understanding how your application functions shouldn’t be so expensive and difficult. Excited to build something that enables developers to provide the best customer experience possible.

[nathanstweeting @nnolantweets]

Starting to build what I hope will be an APM for all, especially indie hackers and small startups. It will be opinionated, low effort implementations, provided spans and analytics, and actual useful metadata. #indiehacker #SoftwareDevelopment

[nathanstweeting @nnolantweets]

I want to provide aggregate data about your application. Keep an eye on important routes, understand how queries perform and understand how libraries and 3rd party dependencies affect your customers.

[nathanstweeting @nnolantweets]

@ylecun Yeah this seems purposefully misleading. Smarter at what? I don’t think LLMs, in their current form, capture the nuance of human problem solving. Can we tokenize human intuition? Can we feed context about the world in real-time to LLMs?

[nathanstweeting @nnolantweets]

The response, by many, to Devin gives “truck drivers should learn to code, autonomous vehicles will be here in no time” vibes. It’s cool and these tools will totally change development. But I think we are far away from speaking actual useful software into existence #AI #Devin

[nathanstweeting @nnolantweets]

Folks with an interest in keeping the AI hype cycle rolling are overstating LLM capabilities in tech. Attention is not all they need to take your job. They need more global context, which you can’t crib from open source. Use LLMs, don’t fear them. #AI #LLMs #Software

[nathanstweeting @nnolantweets]

The instrumentation, discovery, analysis, etc of metrics is a big ask for small and even medium sized teams. Need to lower the cognitive load and barrier to entry #Software #observability #APM

[nathanstweeting @nnolantweets]

Implementing APM tools is difficult, especially for early stage startups and solo entrepreneurs. Might try to build an opinionated solutions that’ll make getting a base level of observability implemented much easier and cheaper.

[nathanstweeting @nnolantweets]

Code isn’t software, it becomes software when it’s running. Knowing how your software works, via logs & metrics is a must. A chef doesn’t serve a pie after only sampling the ingredients. They make sure the finished product is worth serving #code #Software

[nathanstweeting @nnolantweets]

Why is observability such a second thought? We are adamant on tracking test coverage, why not metric and logging coverage? Pushing to main and getting a green build =! a good user experience. #SoftwareDevelopment #DevOps #SRE

[nathanstweeting @nnolantweets]

@mitchsnelson1 @chamath I worked for the only real competitor of EPIC. Like you pointed out, these large EMRs have a lot of say in the industry. They have no incentive to change, I would guess they also lobby to ensure they don’t have to.

[nathanstweeting @nnolantweets]

@PavSoor @chamath And there are already players that have mass adoption. Getting new systems implemented into health systems takes A LOT. A lot of health systems don’t want to invest in the time and money to do it. Unless it’s really worth it.

[nathanstweeting @nnolantweets]

@PavSoor @chamath Let’s say you go to the hospital and stay for a few days. This might result in lab work, prescriptions, maybe some X-rays and billing. All of this is regulated differently, strictly and is interconnected. So having a unified system is preferable

[nathanstweeting @nnolantweets]

@PavSoor @chamath Sorry, just realized I responded to my own tweet instead of yours.

[nathanstweeting @nnolantweets]

@chamath Like the government, healthcare moves slowly. And for good reason, often. But adoption in the broader sense, will be slow

[nathanstweeting @nnolantweets]

@chamath An EMR consists of A LOT of moving pieces that requires consistent integrations. I think supplanting the large EMRs is going to be very difficult. Even carving out little pieces of EMRs will be hard, given integration reqs and current EMR adoption

[nathanstweeting @nnolantweets]

@chamath I worked at one of the 2 largest EMR companies in the US. Legislation has already required health systems to integrate with these large EMRs. So supplanting them will be difficult, even with tech advances.

[nathanstweeting @nnolantweets]

One “hero” on a DevOps/operations team, gone unchecked or even praised, can single-handedly cause so much tech debt. Stop praising people for, “just getting shit done.” When they leave, your organization is screwed. Get it done, but do it right. Think ahead folks.

[nathanstweeting @nnolantweets]

Can we please stop with the overly distributed systems? If you have a hard time determining which thing is messing with your data, it’s time to reevaluate your architecture.

[nathanstweeting @nnolantweets]

@TylerMaran @levelsio You pay for the convenience of the cloud. Managing your own servers will cost you time, but that might totally be worth it.