Signals, decisions and practice from the people building, testing and operating what ships next. Four desks, one question each: what changed, what you have to decide, how it is actually built, and who is building it.
41 stories publishedRead chronologicallyRSS feed
Your AI agent passes demos but fails in production? Motorway's three-layer evaluation framework cut their error rate from 1-in-8 to 1-in-50. Here's the pattern that actually ships.
Four frameworks built in-house at Telerik, a podcast episode about performance testing infrastructure, and a consultancy whose stated aim is to train its clients out of needing it. A useful counterweight to a programme full of agents.
Her published trail is not documentation but hackathons — and the unglamorous business of getting a copilot out of a studio and into Microsoft Teams, where colleagues and their permissions actually are.
He spent thirty years in IT before deciding testing needed orchestrators rather than testers. His open-source Agentic QE Fleet now ships 60 agents across 11 coding-agent platforms — and its own documentation records where the thing got it wrong.
What changed this week, and whether it matters yet.
The call you have to make, and what it costs either way.
The Digital Omnibus just reshuffled your EU AI Act timeline, high-risk obligations pushed to December 2027, but Article 50 disclosure hits in nine days. Here's what actually matters right now.
Thirty-seven days until the original August 2026 deadline, and the Omnibus delay isn't law yet. Here's what still applies and what your team should tackle this sprint.
The EU AI Act's penalty framework looks terrifying on paper – up to €35 million or 7% of global turnover for the worst violations. But the actual enforcement math tells a different story, one where SME protections, mitigating factors, and GDPR precedent suggest most Bulgarian tech companies face far lower exposure than the headline numbers imply.
AWS and Cisco just automated security scanning for MCP servers and AI agents, solving the visibility nightmare that's been keeping security teams up at night. Here's what this actually means for your growing agent sprawl and compliance headaches.
Cypress AI has moved from experiment to production-ready with natural language test generation and self-healing selectors. For teams drowning in test maintenance debt, this represents a concrete decision point rather than another vaporware announcement.
How it is actually built, run and paid for.
Most recent
The first EU AI Act harmonised standard hits publication in August 2026, but your compliance runway is shorter than the calendar suggests. Here's what the Digital Omnibus delay actually means for your implementation timeline.
A 1,000-sample protein design campaign takes 375 hours on 4 GPUs. AWS's BoltzGen-SageMaker integration finally solves the GPU lifecycle headache with per-second billing and step caching.
AWS just shipped a container that turns SAP's OData APIs into first-class tools for AI agents, deployed on a managed runtime with session isolation and OAuth 2.0 out of the box. For Bulgarian teams running SAP workloads, this is the first production-grade path to agentic automation without custom integration code.
Only 7.4% of organizations run unified observability, while nearly half juggle three separate stacks. The problem isn't missing features,it's configuration friction that burns entire sprints on YAML archaeology.
KEDA just got a reference implementation for GPU autoscaling that solves the problem quietly burning money across every Kubernetes cluster running inference workloads. Teams can finally scale on what actually matters: GPU utilization, VRAM usage, and power draw instead of idle CPU metrics.
With 74 days until EU AI Act high-risk enforcement hits, over half of organizations still lack basic AI system inventories. Here's your sprint-by-sprint compliance checklist to avoid €15 million penalties.
The EU AI Act's high-risk system deadline just shifted to December 2027 after the May 7 Omnibus deal. The compliance infrastructure wasn't ready, but this isn't a reprieve - it's acknowledgment that teams need to start building now.
The people and companies in this ecosystem.
Banks are automating model documentation while a visa rule threatens to choke the talent pipeline building those models. The timing is almost poetic, and Bulgarian teams should pay attention to both.
Experian's $50,000 partnership with sports journalist Taylor Rooks isn't just influencer marketing – it signals how data companies are rebuilding consumer trust through authentic voices and transparent products. For Bulgarian fintech teams, this shift toward consumer-first approaches offers a blueprint for DORA compliance and competitive advantage.
Experian's Agent Trust framework introduces Human-to-Agent Binding for AI transactions, creating a new compliance layer that didn't exist six months ago. Bulgarian engineering teams building agentic systems need to track this emerging KYA standard now.
Experian's Small Business Index held steady at 49.2 in March despite inflation jumping to 3.3% and fuel costs spiking 36%. For engineering teams serving small business clients, this stability masks underlying stress that will affect payment cycles, feature priorities, and churn patterns in the months ahead.
The tags carrying the most stories right now. Each one is its own page.
What gets written up here is what gets argued about on stage — and every past edition stays online.
ISTA 2026 - The Future of Technology Conference
Editions from 2019 to 2025 stay online at stable URLs, with the speakers and sessions each one published.