Stop Treating the Model Like a Chatbot
Production AI is more than a chatbot. Models work best as components within larger systems, using structured outputs, search, APIs, databases, and other tools to get the job done.
Production AI is more than a chatbot. Models work best as components within larger systems, using structured outputs, search, APIs, databases, and other tools to get the job done.
A successful AI demo proves a system can work. Real evaluation measures whether it consistently works better by testing correctness, speed, and cost against repeatable benchmarks.
Hinglish, Punjabi-English, Spanglish, Taglish. Hundreds of millions of people mix two languages inside a single sentence — and most AI systems quietly fall apart when they do. Here's what breaks, and what we do about it.
Sometimes good engineering means walking away from something you built well. A real-world build-versus-buy story about evidence, sunk cost, engineering judgment, and knowing when simpler is better.
by Brian Pollack BC, Alberta, and Manitoba gave the clock change the most polite Canadian middle finger. On Nov 1, many of their computers fall back anyway. The Rules Changed Faster Than the Systems Here's why. NTP, which keeps computer clocks in sync, only knows UTC, not time
What happens to our work after we leave? Scheduled blogs, documents, automations, and processes can keep our presence alive long after we've moved on.
AI applications can fail even when the model works perfectly. Real production failures show why timeouts, monitoring, logging, performance, and ordinary software engineering still matter.
by Brian Pollack I feel like it's inevitable that I will get hacked. Citizen Lab just proved spyware can arrive through an online ad without a single click. An Israeli company posing as an ad firm let spyware vendors aim ads at specific people. At least 45 computers
Pulling line items out of supplier invoices and work orders is the most requested AI job we see. We built it three different ways. Here's what each one wins, what each one costs, and how to tell which is yours.
A clever AI search architecture worked exactly as designed - until newer capabilities made it unnecessary. A real-world lesson in how yesterday’s expertise can become today’s complexity.
by Brian Pollack It's all over the news: no more Adobe licenses needed. This is much bigger news than "Claude, make me ph0tosh0p." The truth is wild and the repercussions will be extensive. When Expensive Software Is No Longer Irreplaceable The claim is simple: open source
AI may not have forgotten your instructions. They may be stored in the wrong place. Why persistent project rules need more than a conversation to survive.
AI
When an AI agent gives the wrong answer, the model may not be the problem. A real debugging story shows why developers should check what the model actually received before rewriting the prompt.
AI
The blog about "The Save Button That Doesn't" missed its own scheduled publish date. You can't make this stuff up.
Brian Pollack
by Brian Pollack If you needed to build a wall, would you hire a young go-getter with lots of energy and muscles or a wise mason with 30 years of experience? Honestly, you need both. Tech teams especially need a mix of senior developers and leads who have seen all
Brian Pollack
by Brian Pollack There is a new model every other day, and there is a reason the pace keeps picking up. The models are now improving themselves. The industry term is RSI, or recursive self-improvement. Google has been the loudest about it, but every major lab is running some version
technology
Progress bars look precise, but the percentages aren't always what they seem. A look at how software uses progress indicators to communicate status, manage expectations, and quietly shape what we trust.
AI
Illustration of how ChatGPT and generative AI process conversations as successive interactions rather than experiencing continuous time, using stepping stones and a river of lost context to explain AI context loss and conversational memory.
White Paper
by Prateek Sharma, AI Engineer Almost every AI trial we see does what it promised in the meeting room. Most still never reach a real customer. The gap is rarely the technology — it's five things nobody scoped at the start. Download a pdf of this white paperProtovate_Why_
Brian Pollack
by Brian Pollack Do they think WiFi comes from the actual clouds? The internet needs a new PR firm. Let me explain. If you're posting about data centers while binge-watching Netflix with Reels scrolling by on your phone, you might want to learn where WiFi goes to sleep
AI
AI doesn’t always forget instructions. Sometimes the context window gets overloaded. Two very different AI workflows reveal the same problem - and why managing context matters.
White Paper
by Prateek Sharma, AI Engineer From writing your first effective prompt to building real AI systems: a practical roadmap for learning prompt engineering, and an honest look at what the career has actually become. Download a pdf of this white paperProtovate_How_to_Become_a_Prompt_Engineer.pdf3 MBdownload-circle What
Brian Pollack
by Brian Pollack Nobody hands out cordless saws at orientation. Think about the building you work in. Somebody designed the way people get into it. There is one main entrance where visitors sign in, badge readers on the doors that matter, cameras on the parking lot, a fire marshal who
AI
After being laid off, I’m helping build an AI version of myself to preserve decades of institutional knowledge. But how much can AI capture before knowledge gives way to judgment?