Digest Email: Archive

Date sent: 26 August 2026

Voice AI is moving from “AI phone receptionist” demos to real business workflows: healthcare, KYC, collections, booking and support.

I picked up another ASR model training project and joined a new `Voice AI` global community, which made their first meetup in Toronto.

Watching what researchers can reconstruct from a fruit-fly brain made me wonder.
…or just way too many animated 3D deepfakes?

Date sent: 27 July 2026

AI agents can now see a browser/desktop, click buttons and complete workflows.
Which means QA automation may finally move beyond writing Selenium scripts.

But, I benchmarked a deterministic workflow against a harness agent for non-coding automation & QA. Same task, roughly 20× difference in speed. Agents are powerful – but not every workflow needs an agent.

AI-native moved my mobile game forward, for that project (I started in 2015) I’ve just opensourced MCP for cheaper OpenGL testing.

Date sent: 25 June 2026

Testing AI agents is not the same as testing software.
There may be 10 valid ways to complete the same task — so “expected steps” no longer work.

Job applications are becoming an AI-vs-AI game.
Candidates use agents to apply. Companies use AI to filter applications.
Greenhouse calls it an “AI doom loop.”

I become someone who can walk into a company, understand the workflow and actually deploy AI into it.

Date sent: 28 May 2026

Enterprise problem: employees installing agents such as OpenClaw and Claude Code that can autonomously execute tasks, modify code and access company data

The interesting question is no longer “will developers use AI?” It’s how many developers one AI-native team will need.

S&P goes up, when will the AI bubble burst?

Date sent: 23 April 2026

Codex usage jumped from 3M to 4M+ developers per week in just two weeks.

Snap cut 1,000 jobs while saying AI already generates 65%+ of its new code.

AI is no longer just helping developers write code. Companies are starting to use it to justify smaller engineering teams.

From my side I’ve created an opensource AI-pipeline to apply for job – let’s see who’s who!

Date sent: 26 March 2026

PostgreSQL is becoming the default database for AI projects.

Classic DevOps pitch: “you need ELK, Terraform, migration, full redesign…”
Reality: client just needed a Plesk reseller on their web server.
Be careful — overengineering is expensive.

Hundreds of AI QA tools exist, but teams still waste time fixing broken tests, managers still lack visibility.
We’re now working on our Selenium / Playwright AI layer — looks promising…

Date sent: 26 February 2026

What Businesses Should set up this month (Playbook): MFA, policies on data processed, AI agents (shift from ChatGPT), third-party/vendor access audit.

WordPress attacks are still very real in 2026. Outdated plugins = open door. Remind your clients: update often, avoid random vendors.

We deployed our chatbot testing platform in a bank — team loved it 👏. Finally, visibility into what bot actually does (and break).

Date sent: 29 January 2026

OpenClaw is blowing up the Internet! We are not behind. But before large-scaling, let’s explore how it can be easily on tokens.

Which SDD framework to pick: Spec Kit, OpenSpec, BMAD, GSD, LeanSpec, Traycer… Everyone claims structure, none survive real projects as per our observation.

We released our chatbot testing tool 🎉. Still have spots for closed beta — send your bot if you want real evaluation.

Date sent: 25 December 2025

EU AI Act enforcement phase starts → companies must prove compliance (risk levels, explainability, logging). AI is now a legal topic, not just tech.

Langfuse — ugly UI, powerful backend. Big plus: on-prem, no need to leak your AI data to the cloud. Clients are already asking.

We’re finishing our chatbot testing tool: Domain Competence, Safety, Privacy, Hallucinations, Jailbreaking, Bias, Red Teaming — all covered.

Date sent: 27 November 2025

Ransomware-as-a-Service (RaaS) Reaches Industrial Scale. First Autonomous AI Cyber Attack recorded in China.

Writing a solid article ≠ just using ChatGPT Deep Research. Real pipeline: sourcing → filtering → structuring → writing → editing → publishing. ~6 hours, ~15 AI steps, and still manual work. – How do you create your content?

We evaluated chatbot testing tools: DeepEval, Voltagent, RAGAS, Promptfoo, Arize, OpenLIT, MLflow… none cover even 50% of banking requirements.

Date sent: 30 October 2025

SaaS is slowing as companies shift toward all-in-one AI platforms and on-prem.

Bluetooth beacons are still alive: retail, personal offers, VIP attention. Old tech, but still makes money. – We’re integrating into mobile app.

Banks still run COBOL in prod. Nobody fully understands it. AI helps… but “rewrite to Java” in Claude Code? Doesn’t work. You need a multi-step pipeline with approvals, reviews, and a lot of bureaucracy. – A new big challenge for us.

Date sent: 25 September 2025

LLM costs are exploding → companies start tracking every token.

LLMetry is becoming a must-have: logging LLM usage, A/B testing prompts. Big players like Dynatrace, Elastic, New Relic are entering… but we have discovered their tooling are still raw 😢.

Our AKS insight: Istio routing is ~2x cheaper than Azure Application Gateway in real setups. Not obvious until you see the bill.

Date sent: 28 August 2025

Despite tech layoffs elsewhere, job postings for senior workers (10+ yr XP) in AI, cybersecurity, and cloud have increased by 17% – our recruitment team reports

Claude Code quietly beat Cursor + Copilot in our own vibe-code tests.

Enterprise sprint planning is still a nightmare: Confluence → Jira “messing around” → 15 repos → release plan… Even estimating ONE ticket is chaos. We’re trying AI: pull Jira + Confluence via API, chunk/embed, analyze… sounds easy until it doesn’t 😅