AI is a work in progress. We make it work for you.

Joined July 2025
This week, we went to @IFS Unleashed to announce IFS Intelligence. Zach, our eng lead on the project, chatted with Harshana from IFS about building it as one team across two companies. More to come.
1
2
1
12
1,035
We extracted nearly half a million business rules from the code behind Medicare claims processing. CMS experts graded random samples each week like homework. That review was one of seven stages in the pipeline. Zachary Gonzalez walks through the process in our Modernizing Legacy Systems webinar.
1
9
842
Watch the full seven-stage walkthrough: 8090-solutions.link/x-modern…
171
GLM’s cheaper runs cost more per success. Codex: $0.11/run, 24% passed. Pi: $0.13/run, 58% passed. Including failed runs, API cost per success was $0.46 in Codex and $0.22 in Pi. Costs use Z.AI’s promotional rate, which ended Sept 9. At list rates, both double.
1
7
873
Read the findings and inspect every prompt, result and agent trajectory: 8090-solutions.link/x-same-m…
206
Before an AI agent touches a legacy system, we map what that system can do. That map is a catalog of actions. Each one is linked to when it applies and which permissions govern it, built from three sources: the old code, the vendor systems you rent, and the experts who have run the system for years. People and AI agents then work from the same map, through one governed gateway like MCP.
11
7
1
102
73,913
Aditya Bhatnagar from our technical staff walks through how it comes together in our Modernizing Legacy Systems webinar. Watch the full webinar: 8090-solutions.link/yt-moder…
1
5
2,026
Changing only the harness moved GLM 5.3 Flash’s coding success rate from 24% to 59%, a 35-point gap. 8090 Labs tested 7 models in 5 harnesses on @datacurve's DeepSWE v1.1, running each of the 113 tasks twice per pairing. Here’s what we found:
3
2
1
28
282,732
We ran each task twice to check whether a pass repeated. GLM passed both runs on 51 of 113 tasks in the DeepSeek harness, compared with 16 in Codex. Two runs per task give an initial view of repeatability, but more are needed to establish reliability.
1
4
338
Test the model and harness together, then check whether the results repeat. Every prompt, result and agent trajectory is public: 8090-solutions.link/x-same-m…
203
We’re live tomorrow at 9 AM PT / noon ET. Join Aditya Bhatnagar & Zachary Gonzalez for a walkthrough of modernizing COBOL and Assembly systems. Register: 8090-solutions.link/x-modern…
Join us for a live session on modernizing COBOL and Assembly systems with Aditya Bhatnagar & Zachary Gonzalez from our technical staff. We'll cover how legacy modernization works across the public and commercial sectors, then walk through our business rules work on CMS's ClaimsCore program as a case study. October 7 · 9 am PT / noon ET Registration below.
1
2
24
21,383
modernization, auditability, control, evidence...
Join us for a live session on modernizing COBOL and Assembly systems with Aditya Bhatnagar & Zachary Gonzalez from our technical staff. We'll cover how legacy modernization works across the public and commercial sectors, then walk through our business rules work on CMS's ClaimsCore program as a case study. October 7 · 9 am PT / noon ET Registration below.
36
8
155
83,131
Mentix went from a conversation with a surgeon to a working version in about three weeks. Utopia Studio built the surgical mentoring platform using Software Factory, starting with conversations with Dr Amyn Haji, a consultant surgeon at King’s College Hospital. He wanted doctors anywhere in the world to get live advice from an expert during a procedure. Defining the product meant understanding how surgeons learn, including the physical feedback they feel and the cues they read in theatre. The team captured that knowledge in Software Factory and used it to create Requirements and Work Orders. A founding engineer joined with no prior project context. He pulled the Work Orders into a coding agent and started building, with the surgeon’s requirements and the decisions behind the work already documented.
1
1
7
1,123
The full story of building Mentix with Software Factory: 8090-solutions.link/x-utopia…
373
To CEOs, CFOs, CIOs and Directors: Have you read the White House Accord on Super Intelligence? What will you do to safely deploy Super Intelligence within your Enterprise? EY and 8090 are building a new solution - KYSI (Know Your Super Intelligence) designed to create a verifiable record of AI-enabled decisions by capturing the context and evidence, and spotlighting controls, across enterprise environments. The solution helps organizations strengthen governance while keeping pace with innovation. It does this by providing: - End-to-end traceability - Policy-to-risk mapping - Auditable evidence packages Together, we are helping organizations move beyond AI experimentation toward AI they can trust, govern and scale enterprise-wide with confidence. To learn more, reach out to your EY Partner or ping us at 8090.ai
101
87
9
636
168,870
Rebuilding a legacy system means preserving business rules buried in decades of code. Join our October 7 webinar with Aditya Bhatnagar & Zachary Gonzalez to see how we use AI to extract those rules, verify them with domain experts and use them to rebuild the system. Register: 8090-solutions.link/x-modern…
1
8
824
Evolve with 8090. We're opening the doors to our Redwood City office for EVOLVE, an overnight hackathon for experienced engineers to build AI agents and tools that modernize legacy software. Apply: 8090-solutions.link/x-evolve
1
6
5
21
4,125