Blitzy

Blitzy set another record on SWE-Bench Pro with a score of 84.95%, surpassing our previous leading score from March (66.5%). In building Blitzy, we've seen firsthand that the system surrounding a model is far more important than the model itself. These records demonstrate the value of an abstraction layer and agentic orchestration for the enterprise. By allowing a platform to orchestrate multiple model providers it derisks an agentic SDLC by lowering the cognitive burden on engineers and lowering the reliance on any given vendor. Both results were independently verified by Quesma Our most recent Blitzy blog explains the technical decisions that drove this achievement. Read here: https://lnkd.in/gSEjHnJF

SIs charge millions and take a year to document a legacy codebase. Blitzy does it in days, for free. It ingests the code and builds a self-improving knowledge graph of every dependency, abstraction, and architectural decision buried inside. The institutional knowledge of a principal engineer who has been there for decades. You don't need to pay someone to figure out how it was built. Blitzy already has. Learn how at blitzy.com

In Part 2 of Measuring Engineering Velocity in the Age of AI, Jellyfish AI Advisor Adam Ferrari and I go beyond how enterprises measure AI impact to what's actually possible when you rethink the SDLC from the ground up. Here's what the autonomous SDLC unlocks. "Everything from design, implementation, testing, security, and QA - you can deploy agents, orchestrate them autonomously, and do the bulk of the work. And when you've done that, we are seeing 500%, 5x velocity increase. I think the world is yet to even scratch the surface of the potential of something like this." Full conversation on YouTube. Links to Part 1 and Part 2 in comments.

Strong reverse engineering directly translates into greater throughput. However, this isn’t the only important factor for accelerating your software development. For regulated industries like insurance, generating high-quality and deterministic code is also top of mind. Kris Tokarzewski, Vitality Chief Technology & Information Officer, was on CXOTalk discussing exactly how Blitzy delivered deterministic output and reverse engineered their legacy systems. Listen to the full conversation linked in the comments.

Guardrails are a critical component of input specifications. We built Blitzy with this in mind. Blitzy’s input covers more than project objectives and tech specs. The platform’s prompts also account for security considerations and an organization’s specific guidelines. Validation then happens throughout the development process. This level of detail directly reflects the rigor required for enterprise engineering. GNP Seguros CIO & Head of Transformation Enrique Ibarra sat down on CXOTalk to discuss: - Results achieved: Increased 5-10x engineering velocity and 80-95% task completion - Adding security, governance and architectural guardrails inside prompts - How Mexico’s largest insurance company used Blitzy across four distinct use cases Full conversation linked in comments.

Reverse engineering legacy systems just happened 5x faster. Vitality Chief Technology & Information Officer Kris Tokarzewski was on CXOTalk discussing exactly how Blitzy delivered deterministic output and reverse engineered their legacy systems. This conversation highlights the significance of reverse engineering and determinism compared to probabilistic output for regulated industries. Link to the full episode in the comments.