Otherwise, people will just build custom tools for themselves.
If AI can generate software, infrastructure becomes the bottleneck.
Not AWS-scale cloud but simple, cost-capped, AI-accessible infra for small, single-user apps.
Feels like an inevitable shift.
(1/1) Next big thing is micro infra. Let me tell you what I mean:
Coding is becoming cheap and abundant. When everyone can build apps easily, code is no longer the moat. The real differentiation shifts to:
• Problem selection
• Cost efficiency
• Operational simplicity
Daily mess of traffic near Ryan Public College Vasant Kunj.
Standing here for 30 minutes and I am not able to move even 200m.
Why? No one care about the red light.
BTW its around 8:30 in the morning.
@dtptraffic
For anyone still thinking Software Engineering would remain as a job rather than an additional skill.
Use @claudeai Opus4.6
You will see you career plans fading away.
AI summit made headlines.
Not for the ideas or innovations, but for the drama around it.
That says less about AI and more about us as a society.
What we choose to amplify, watch, and share.Serious work still exists.
You just have to look past the noise to see it.
#IndiaAISummit
Mukesh Khanna is back as "Shaktiman"! 💪 🦸♂️
But seriously, is social media making our celebs delusional? 🤔People love to see praises and the algorithm loves to feed us feel-good content, and if a celeb sees a lot of praise, they might start believing their own hype.#shaktimaan
Talking more about videos or contents that are more used in making summaries. For example podcasts, educational videos etc.
A particular example from a video may connect with you more than another person, therefore should be highlighted in your summary.
I am not sure if I am thinking in the right direction,
But next evolution of video summary would be personalised based on user.
I believe everyone of us is different and we pick up different things and remember different things from a content.
Things to keep in mind when building LLM systems. 😷
Below are a few mentioned LLM limitations that can give you a cry if you haven't thought of it before building a system around it:
12/ LLMs need access to context and data. Ensure this doesn't compromise user privacy. Consider self-hosting LLMs or maintaining user personalization without exposing sensitive data. Robust security measures are paramount, including encryption for data and secure APIs
Demystifying LLM Systems
1/ Building the future systems with #LLMs? Hold on! There's more to it than just the models themselves. #AI#MachineLearning#LLM
11/ Combining multiple LLMs with various tools can slow down response times. Consider building an LLM cache - a repository of previous outputs that improves efficiency for common queries. This cache can be enriched with user feedback for ongoing optimization. #LLM#AI