Year in Review: 2018 in Tech
2018 was a year of maturation. The hype cycles continued, but practical applications caught up. Here’s my take on the year in tech.
AI/ML: The BERT Moment
BERT was the headline. Google’s pre-trained language model crushed benchmarks and fundamentally changed how we approach NLP. Transfer learning isn’t new, but BERT made it practical for everyone.
Beyond BERT:
- AlphaFold showed AI could tackle protein folding
- GPT demonstrated generative language capabilities
- AI ethics moved from academic concern to industry focus
The shift: From “can we build it?” to “should we build it?”
Cloud: Multi-Cloud Reality
AWS remains dominant, but Azure grew 76% YoY. Google Cloud is finding its niche with Kubernetes and ML.
The industry accepted that multi-cloud is inevitable:
- Avoid vendor lock-in
- Regulatory requirements
- Best-of-breed services
Kubernetes became the common denominator. If you’re not on K8s, you’re explaining why.
JavaScript: The TypeScript Tipping Point
TypeScript adoption accelerated. Major frameworks (Angular, Vue 3) embraced it. Developer confidence in large codebases improved dramatically.
2018 also gave us:
- React Hooks announcement (preview)
- Vue 3 planning (Composition API)
- Node.js hitting 10 LTS
The trend: JavaScript grew up. Tooling for large-scale development became the norm.
Python: 3.7 Lands, 2 Fades
Python 3.7 brought dataclasses, and the community finally, finally started abandoning Python 2.
Django 2.0 went Python 3-only. Popular packages dropped 2.7 support. The writing is on the wall—January 2020 is coming.
DevOps: Platform Engineering Emerges
“DevOps” became too broad to be useful. Platform engineering—building internal developer platforms—emerged as a focus area.
Key themes:
- GitOps with tools like ArgoCD and Flux
- Service meshes (Istio, Linkerd) for complex networking
- Observability as a discipline (Prometheus, Grafana, Jaeger)
The realization: Developer experience is a force multiplier.
Hardware: The ML Chip Race
Custom silicon for ML training/inference became a battleground:
- Google’s TPU v3
- NVIDIA’s Turing architecture
- Intel acquiring Nervana
Edge AI started emerging as a category. Running models on-device isn’t just a demo anymore.
Security: Data Breaches Escalate
GDPR went into effect. Companies scrambled to comply.
Meanwhile, breaches continued:
- Marriott: 500M records
- Facebook: 50M+ accounts
- Quora: 100M users
The pattern: Collect data, get breached, apologize, repeat.
Open Source: Licensing Battles
Redis, MongoDB, and others changed licenses to prevent cloud providers from offering managed versions without contributing back.
The debate:
- Open source sustainability is a real problem
- But restricting use feels against the spirit of OSS
No clear resolution. Expect more license innovation.
My Predictions for 2019
High confidence:
- Python 2 death creates migration pressure
- Kubernetes becomes table stakes
- TypeScript adoption accelerates
- BERT derivatives proliferate
Medium confidence:
- Serverless goes mainstream for specific use cases
- Edge computing gains traction
- AI regulation discussions intensify
Low confidence:
- WebAssembly breaks out beyond niche use
- Blockchain finds non-speculative utility
Personal Highlights
Technologies I adopted in 2018:
- Used heavily: Kubernetes, TypeScript, Prometheus
- Experimented with: GraphQL, Rust, Terraform
- Abandoned: Redux (for simpler state management)
- Watching: WebAssembly, Deno
Lessons Learned
- Boring technology is underrated. PostgreSQL, Python, Linux—they just work.
- The FAANG papers are worth reading. BERT, TensorFlow papers informed real decisions.
- Platform thinking pays off. Investment in internal tools multiplies developer productivity.
- AI ethics can’t be an afterthought. The industry is waking up, slowly.
Looking Ahead
2019 will likely be:
- More AI in production (not just research)
- Cloud consolidation
- Continued JavaScript fragmentation then reconsolidation
- The Python 2 funeral
Technology moves fast. Fundamentals move slowly. Focus on fundamentals.
See you in 2019.