TL;DR — In 2025 we shipped ten production releases across client work and our own products, expanded Blink AI’s coaching engine, launched Wallnetic on the Mac App Store and grew Nexia Academy to multi-tenant deployments. Here is what we built, what we got wrong, and what we are carrying into the rest of 2026.
Year-end posts are usually marketing fiction. This one is engineering arithmetic — written with a few months of distance from the year it covers, because shipping product releases edged out blog drafts in Q1. T-Square has run as an independent software studio since 2017, and 2025 was our most product-heavy year so far. Ten production releases across client engagements and our three in-house products. Thirteen brand engagements. Eight industries.
Below is what we shipped, where we made trade-offs and the failures we cared enough about to write down.
Products: three running in-house
Blink AI — our AI-powered corporate coaching product built around the GROW methodology and Socratic questioning — gained a new prompt orchestration layer, a redesigned admin dashboard for L&D teams, and tighter latency on the conversation API. We added retrieval-augmented memory so a coachee’s history persists across sessions without leaking context between cohorts.
Nexia Academy — our white-label learning platform — moved from a single-tenant deployment to a true multi-tenant architecture with isolated schemas, per-tenant branding and per-tenant SCORM/xAPI exports. Three new tenants went live in 2025.
Wallnetic — our native macOS live wallpaper engine — launched on the Mac App Store. The submission process taught us more about Apple’s sandbox enforcement and entitlement edge cases than two years of iOS app work combined.
Client work: ten production releases
Across 2025 we shipped ten production releases for clients — corporate training platforms, AI integrations on existing products, custom backend services, mobile apps and DevOps modernization projects. Every release passed through senior engineering review, ran in staging for at least one full sprint and went live with on-call coverage from the team that built it.
What we got wrong
- Over-indexing on Postgres + pgvector early. For one product we should have moved to a dedicated vector store at 5M embeddings, not 12M.
- Under-investing in observability before scale. Two incidents in Q2 would have been five-minute fixes with the OpenTelemetry stack we deployed in Q3.
- Treating cmi5 LMS support as documented. Three LMS vendors claimed cmi5 support; only one shipped a working launch flow. We now validate end-to-end before promising features.
What is playing out in 2026
The 2026 roadmap is already in motion: deeper AI integration in Nexia Academy, a public API for Blink AI in progress, and a second native macOS product following Wallnetic’s pattern. Engineering-wise we have standardized on TypeScript across the full stack where possible, expanded OpenTelemetry coverage, and continue to ship under KVKK and GDPR-aligned data-processing addenda.
Frequently asked questions
How many engineers work at T-Square?
T-Square operates as a senior-only studio — every commit is written and reviewed by engineers with at least 8 years of production experience. There are no juniors hidden behind invoices and no offshore sub-contracting.
What products does T-Square run in-house?
We operate three production products: Blink AI (AI corporate coaching built around GROW methodology), Nexia Academy (white-label learning platform), and Wallnetic (native macOS live wallpaper engine on the Mac App Store).
Working on something similar?
T-Square is an independent software engineering studio. We architect, build and operate production-grade systems for learning, AI and custom software products. Talk to a senior engineer if you’d like a second opinion on your architecture or roadmap.