Blitzy raises $200M at $1.4B valuation to autonomously refactor enterprise legacy codebases over weeks

Blitzy is announcing a $200 million financing round at a $1.4 billion valuation. The company builds an autonomous software development platform aimed at enterprise legacy codebases — banks, insurance companies, and similar organizations sitting on tens of millions of lines of code.

What Blitzy actually does

The core pitch is duration. Most autonomous coding tools, Elliott argues, run for thirty minutes to a day before handing back to a human. Blitzy is built to run for days to weeks, continuously and recursively improving code using models from OpenAI, Gemini, and Anthropic in combination. Within 48 hours of installation, the system reverse-engineers a codebase — often 30 to 50 million lines — and builds a knowledge graph that agents can traverse to understand relational dependencies across the code before touching anything.

That relational framing is the technical crux. Elliott's view is that code isn't serial — you can't just feed 100k-token chunks to an LLM in sequence and expect coherent output on a large codebase. Something relevant to one service may live in a completely different part of the stack. Blitzy claims to have built a language-agnostic, version-agnostic way to model those relationships, which is where it argues competitors fall short.

“Blitzy, we're announcing the $200M financing at a $1.4B valuation today. We're an autonomous software development platform specifically designed for complex enterprise use cases — banks, insurance companies, anyone with huge amounts of code. The system will run for days to weeks autonomously, recursively improving the code using all the foundational models together.”

Context window limits

Elliott is direct that advertised context windows beyond roughly 100k tokens are largely illusory in practice. Models use sparse attention beyond that threshold, and intelligence degrades quickly — what academics call "context pressure." Throwing a slightly larger context window at a 50-million-line codebase doesn't solve the problem. A system-level approach to what the model sees, and when, is what matters.

Go-to-market

The sales motion is direct enterprise, described as Palantir-like: go in, demonstrate value fast, and work through complex brownfield environments that off-the-shelf tools can't handle. The "build vs. buy" question is live in every deal, but the 48-hour reverse-engineering demo is apparently the wedge.

Competitive positioning

Elliott's positioning argument is that "autonomy" has been defined down by the market. Running for a day counts as autonomous to most players. Blitzy is trying to set the benchmark at weeks of continuous, inference-heavy operation. Whether that definition holds as models improve is the open question, but Elliott's framing is explicitly that better foundation models make Blitzy more capable, not less relevant — the company calls itself "long transformers" and wants as much industry CapEx into AI compute as possible.

The company's near-term bet is that no competitor is as narrowly focused on large-scale brownfield autonomy, and that focus compounds as model quality rises.