Your trusted AI software development partner
San Jose, CA
Joined May 2014
- Tweets256
- Following14
- Followers116
- Likes54
"It's not models, it's always harness."
Our CEO Vadim Peskov on the @GoodFirms podcast: why AI coding tools speed up typing but not teams, and what delivers 5-30X on real systems.
Coding was never the bottleneck in software. Deciding what and how to build is.
On the @GoodFirms podcast I explained why AI coding tools speed up developers but rarely speed up teams, and why the harness matters more than the model.
youtu.be/mzIGQUUgCYY
Diffco has been named a Top 10 IoT Development Company for Q3 2026 by @50Pros! 50pros.com/agency/diffco
Diffco retweeted
Check out my latest podcast on how #AI engineering is changing the software development lifecycle and what is actually important for building software with AI youtube.com/watch?v=2bhcz_LB…
Diffco retweeted
Soon you will be to can generate movie in real time and stream different view to each viewer
Today is a very historical moment for AI video generation
You can now generate AI video faster than you can watch it
Before it'd take let's say 2-5 minutes to generate 15 seconds of video
@fal made a post-trained Minimax H3 variant called Max which is 50x faster than the original but still maintains quality
It generates 15 seconds of video in 9 seconds!
That means you can now do new things like build a perpetual livestream with it that never ends!
Diffco retweeted
Better models don't mean better software.
I went on @PaleBlueNexus to explain why — after 18 years of building software and a year of building it agentically.
4x cost reduction. ~100% AI-written code on new projects. And the one rule that makes it work 🧵
youtube.com/watch?v=4Zcf0qmP…
Diffco retweeted
Multi-agent support system teardown: 4 iterations, 58% → 87% resolution.
We didn't start at multi-agent. Single-agent failed first, and we earned each layer.
diffco.us/blog/multi-agent-s…
AI Self-hosting math people skip:
A GPU fleet sized for peak runs at 30–60% utilization → effective cost 1.5–3x the headline rate.
Break-even is usually sustained hundreds of millions of tokens/day. Below that the API wins.
diffco.us/blog/self-hosted-v…
Our hard rule for AI in regulated finance:
Models propose. Deterministic services execute. The model never holds the keys to a money-moving API.
Policy gates are unit-tested code, not prompts. The model can be wrong. The gate can't.
diffco.us/blog/ai-in-fintech…
Diffco retweeted
AI-native vs AI-assist development: two different jobs, often confused. Different skills, different hires, different failure modes. Stop merging them. Check our blog here:
diffco.us/blog/ai-native-vs-…
Diffco retweeted
A vibe-coded product typically lasts 3–9 months before the foundations crack. The answer isn't "stop vibe coding." It's "don't vibe code forever." One engineer in the first 60 days costs a fraction of the rebuild at month 12.
diffco.us/blog/why-you-still…
Check out our interview about: How Digital Agencies Build and Deliver norobots.co/how-agencies-bui…
Check full post here diffco.us/blog/vibecode-resc…
2/ More:
— References are demos, not customers — They've never said "no" in the sales process — They can't quote cost-per-request and p95 latency on their last 3 builds
If they can't show numbers, they're guessing — and you'll be guessing with them.
3/ The differentiator in 2026 isn't AI capability. Every shop has the same models, SDKs, frameworks.
It's engineering discipline. Eval rigor, observability, cost control.
The model fashion will be obsolete in 18 months. Discipline outlasts the model.
diffco.us/blog/how-to-choose…