.webp)
Codebridge has launched shvydko.ai, a decision-AI product for developers, as investor interest grows in AI tools used for decision-making inside business software.
The product turns supplied text, questions and criteria into structured outputs that applications can use. Developers can use it to determine which category applies to a piece of text, assess how strongly a condition is met, or judge whether a request satisfies stated requirements.
Its API accepts context, questions and criteria, then returns a choice, score or probability. The launch also includes a browser-based Playground, API keys, documentation and integration examples to help software teams test the service in their own workflows.
One sample use case is customer support triage. In that example, an application submits an incoming message with descriptions for billing, technical support and sales, and the system returns the most appropriate destination for the software to use when suggesting a queue.
Codebridge said teams would still need to test edge cases before using any response to trigger an action. Messages may contain several issues, miss important information or fall outside the supplied rules, leaving the surrounding application responsible for permissions, escalation and any action taken.
Comparison data
Alongside the launch, Codebridge published an exploratory comparison with Jev 1.14, a separate product developed by TypeSafe AI. On a Ukrainian development-data panel of 1,662 questions, shvydko.ai recorded 90.50% accuracy under what Codebridge described as an equal-family macro metric, compared with 87.57% for Jev.
The reported gap was 2.93 percentage points, with an adjusted 98.33% confidence interval of +0.77 to +5.10 percentage points. Codebridge said the metric gave equal weight to three task families: entailment, inference from supplied passages and rule application.
It also distinguished that metric from overall question accuracy. The result related to a working shvydko.ai v2 service snapshot measured on 16 October 2026, and Codebridge described the exercise as a team-run diagnostic evaluation on development data rather than an independent benchmark.
Codebridge added that the findings did not establish general superiority over Jev. It also said a broader criterion requiring superiority across all three language slices was not met.
That caveat is significant because comparisons in this category can vary sharply depending on the documents, rules and exceptions used in a particular workflow. Teams considering integration should test representative and difficult cases from their intended use before deciding how to deploy the service.
Market backdrop
The launch comes shortly after TypeSafe AI disclosed a large funding round for Jev. On 9 October, the company announced a USD $870 million Series A at a USD $7.5 billion valuation, led by Andreessen Horowitz with participation from Sequoia Capital and existing investor DCVC.
Codebridge pointed to that funding as a sign of growing interest in software that helps businesses make text-based decisions inside operational systems. At the same time, it emphasised that TypeSafe AI is a separate company and that the funding figures relate to Jev rather than shvydko.ai.
Founders
The people behind shvydko.ai are Codebridge co-founders Konstantin Karpushin and Myroslav Budzanivskyi. Karpushin is Chief Executive Officer and Budzanivskyi is Chief Technology Officer.
Karpushin previously worked at KPMG, where he was a Partner and led innovation and technology work. Budzanivskyi, who now leads engineering and architecture at Codebridge, previously worked at GlobalLogic, KPMG and Cybercube.
Codebridge was founded in 2021, and the founders have also worked on the voice typing product Lispr. With shvydko.ai, the company is adding a product aimed at structured decisions within software workflows rather than voice input.
Codebridge said a structured response makes an answer usable by code, but it does not prove that the underlying judgement is correct.