3 ms·

AI Agents Are Here, but Can Generic LLMs Solve Specific Enterprise Use Cases?

2 points by pavankamat24 2y ago

pavankamat24 2y ago
In mid-last year, while brainstorming what to build and how the market would evolve, we made two broad predictions: 1. Overall cost of software development will reduce: Building software will require fewer people, and the speed to ship software will drastically increase. 2. One size doesn’t fit all: The evolution of LLMs in software will start with generic LLMs, much like how SaaS was born in the 2000s — mostly generic — but will evolve to become more and more specific over time. Slowly but steadily, we see these predictions becoming a reality. More and more agents are becoming the talk of the town, each solving one problem at depth. At the same time, LLMs have been progressing rapidly, becoming stronger day by day. But when put to use for context-specific tasks, they still deliver suboptimal results. LLMs need context more than humans to perform well at any given task. A generic LLM trained on terabytes of broad, non-contextual data, capable of answering any question on earth, is inherently limited when tackling tasks whose context and results are specific to a business. Enterprises are experiencing FOMO. They love LLMs but are frustrated because their LLM doesn’t fully understand their business. Our code review agent was the first step in addressing this gap — making context meet code. With this, we’ve enabled 15+ customers to bridge the gap between their unique business context and the code they build. Today, our code review agent is used by 500+ developers, has reviewed 1.5M+ lines of code, and achieved a developer acceptance rate of 27%. Now, we’re doubling down on enabling this experience of making context meet code, not just for code review but across the entire SDLC. This means building specific language models for large enterprises, trained on their own proprietary data, to execute focused tasks with security at the core. These models will deliver better results by aligning precisely with business needs. We have a new digital home — We’ve revamped our website, watch us build amazing stuff at pantomax(dot)co

2 points by pavankamat24 2y ago

In mid-last year, while brainstorming what to build and how the market would evolve, we made two broad predictions: 1. Overall cost of software development will reduce: Building software will require fewer people, and the speed to ship software will drastically increase. 2. One size doesn’t fit all: The evolution of LLMs in software will start with generic LLMs, much like how SaaS was born in the 2000s — mostly generic — but will evolve to become more and more specific over time. Slowly but steadily, we see these predictions becoming a reality. More and more agents are becoming the talk of the town, each solving one problem at depth. At the same time, LLMs have been progressing rapidly, becoming stronger day by day. But when put to use for context-specific tasks, they still deliver suboptimal results. LLMs need context more than humans to perform well at any given task. A generic LLM trained on terabytes of broad, non-contextual data, capable of answering any question on earth, is inherently limited when tackling tasks whose context and results are specific to a business. Enterprises are experiencing FOMO. They love LLMs but are frustrated because their LLM doesn’t fully understand their business. Our code review agent was the first step in addressing this gap — making context meet code. With this, we’ve enabled 15+ customers to bridge the gap between their unique business context and the code they build. Today, our code review agent is used by 500+ developers, has reviewed 1.5M+ lines of code, and achieved a developer acceptance rate of 27%. Now, we’re doubling down on enabling this experience of making context meet code, not just for code review but across the entire SDLC. This means building specific language models for large enterprises, trained on their own proprietary data, to execute focused tasks with security at the core. These models will deliver better results by aligning precisely with business needs. We have a new digital home — We’ve revamped our website, watch us build amazing stuff at pantomax(dot)co