ProjectThunder (@ProjectThunder) header
ProjectThunder (@ProjectThunder) profile picture

Sana Commerce & BigCommerce partner and customization experts. Do you need a #Mobile App 📱#eCommerce 🛍️ #SaaS solution? DM us today! 🇺🇸✨🌌 @calvinluttrell

Palm Desert, CA
Joined March 2009
Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)Photo by ProjectThunder (@ProjectThunder)
Congrats to our valued client @verabradley for winning two awards with their PTI @sanacommerce e-commerce website. 👏 🏆 🏁
1
4
Carnegie Mellon is putting computer science courses on YouTube for free, per its School of Computer Science. Tools change every year. The fundamentals underneath them barely move. The teams that run integrations well usually have someone who understands how systems talk to each other, not just which buttons to press. A few hours on the basics pays back the next time a sync breaks. Which fundamentals do you wish everyone on your team understood? #CarnegieMellon #ComputerScience
3
"It didn't work" is the whole bug report from the field. Build your app so that's enough. Field users can't describe a sync failure. They shouldn't have to. Add a diagnostic export: 1. One button that packages recent logs and the pending queue. 2. It goes to your support team, it isn't read by the user. 3. It holds enough to reproduce the issue without another phone call. It's part of how we build offline-first .NET MAUI apps. • Everyone works California time, developers included. • Four layers of testing, then you sign off on agreed test cases. • Tested on real iPhones and Android phones, not just simulators. #dotnetMAUI #MobileDev #FieldService
4
If a slice of your migration can't ship for six months, it can't be corrected for six months. Big-bang migrations hide their mistakes until the end. Ship in slices instead: - One service - One screen group - One integration Each slice goes live, gets used, gets fixed. Then run old and new in parallel with real users on the new path before the old one is switched off. • Everyone works California time, developers included. • Four layers of testing, then you sign off on agreed test cases. • The developer who does the work estimates it, and AI checks the estimate. #LegacyModernization #DotNet #SoftwareEngineering
3
Nothing makes a field rep hate an app faster than a logout in the middle of an order. B2B mobile runs on corporate identity like Entra ID or Okta. The pattern that works: OIDC with PKCE, proper token refresh, and a clean re-auth path when a token expires mid-session. Get that wrong and the rep goes back to paper. • One .NET MAUI codebase for iOS and Android, so both behave the same. • We test on real iPhones and Android phones, not just simulators. • Four layers of testing, then you sign off on agreed test cases. How many times a week does your team re-enter a password in a work app? #MobileApps #dotnetMAUI #B2B
1
8
A Groundwork Collaborative investigation says nearly 1 in 3 Amazon Black Friday deals it checked offered no savings. B2B buyers notice this faster than anyone. They already know their contract price. If a promo shows a 'was' price they never paid, trust drops on every order after. Show each buyer their own price and what the deal actually saves them. Does your store's sale price compare against what each buyer actually pays? #BlackFriday #Ecommerce
4
Where AI pays off first in a B2B shop: the work your team keeps retyping. Start where text already exists and a person rewrites it: PDP descriptions from spec sheets. Attributes from supplier feeds. Translations and brand variants. Ticket lookups before a reply. The source is real, the output is checkable, and a human approves it. Save the clever stuff for after these pay. #AI #B2B #Operations
"We run Epicor" is not enough information to scope a B2B storefront. • Everyone works California time, developers included. • 22 years of ERP accounting integration, Epicor included. • We run an ERP in-house and hand ERP developers actionable tickets, not blame. Epicor's own index lists ten ERPs. Prophet 21 extends through an SDK and DynaChange. Kinetic extends through Application Studio and Epicor Functions. They share almost nothing. Epicor Commerce, its first-party storefront, supports Kinetic, Prophet 21, BisTrack and Eclipse, and features differ by ERP accounting system. First question on any call: which Epicor, which version, hosted or on-prem? #Epicor #Prophet21 #B2Becommerce
9
Qwen's Qwen4-Max writeup is the top AI thread on Hacker News: 675 points, 317 comments. Another model is at the top of the charts this month. The teams that benefit are the ones that can swap models without rewriting their app. Keep your prompts, your test cases and your model choice separate. Then trying a different model takes an afternoon instead of a project. When a model like this shows up, how long would it take you to test it against your real workload? #Qwen #AI
1
1
Self-hosting map tiles has gotten cheap and easy, per a Protomaps writeup on Hacker News (130 points). Most B2B storefronts have a map somewhere: dealer locator, branch pickup, delivery zones. That map is often a per-request service with its own invoice. A map file on your own hosting can mean one less bill and one less outside dependency. Does your storefront have a dealer locator, and do you know what it costs each month? #Protomaps #Ecommerce
6
On Hacker News: Claude Code and Codex reverse engineered a 1994 Jeep's ECU over a $3 Arduino. Most businesses have their own 1994 ECU: an old ERP accounting system or warehouse tool nobody can fully explain. AI agents are getting good at reading what old systems actually do. Start there. Have them document and map it, read only, before anyone changes a thing. What's the oldest system in your business that still runs on tribal knowledge? #AI #ClaudeCode
4
1
22
A handwritten dismissal notice about 200 years old was found on a roof tile, per the BBC. Someone wrote it down, and it can still be read two centuries later. Many businesses can't read their own order history from before their last system change. When you move to a new ERP accounting system or storefront, decide early what history comes along and in what format. How far back can you pull a customer's order history right now? #History #Accounting
1
7
PyPI's redesigned project pages with clearer security signals are now live, per the PyPI blog. Good moment to check what your own stack pulls in. Most B2B storefronts and ERP accounting integrations depend on dozens of packages nobody chose on purpose. Know the list, know who maintains each one, and pin the versions you ship. Could your team list every package your integration installs today? #PyPI #Python
3
University of Michigan researchers report generative AI uses more energy on non-English text. If your B2B storefront, product descriptions or support bot run in several languages, don't assume each one costs the same. Measure speed and spend per language before you roll an AI feature out to every market. Have you compared your AI costs across languages yet? #GenerativeAI #AI
2
Jonathan Blow's talk "If AI solved coding, why is my software still the same?" is on Hacker News, 509 comments. For most businesses, the code was never the slow part. The slow part is deciding what the software should do: pricing rules, approvals, who owns the customer record. AI writes code faster. It can't make those decisions for you. Where do your software projects actually stall, and is it ever the code? #AI #SoftwareDevelopment
9
Perplexity published how it trains and evaluates Decider v1.3, and the writeup is on Hacker News. The eval part is the one worth copying. Most teams adding AI to quoting or order entry never write down what a good answer looks like. Pull 20 real questions from your order desk and score every model change against them. What would be the first question on your list of 20? #Perplexity #AI
6
If your app can't show what hasn't synced yet, your users will tap Save five times. Offline-first means the user always knows the state of their work. Three cues we build into field apps: 1. Saved on the device, waiting to send. 2. Sent and confirmed by the server. 3. Needs attention, with a reason. Local write, server confirm, clear pending state. Users stop guessing and stop entering things twice. • Everyone works California time, developers included. • One .NET MAUI codebase for iOS and Android, so both behave the same. • Tested on real iPhones and Android phones, not just simulators. #dotnetMAUI #MobileDev #Offline
5
Can't update the callers of your old SOAP service? Don't wait for them. When a partner or a device sits on the other end of a legacy service, you don't control their schedule. So don't tie yours to it: 1. Build the new API. 2. Keep a thin compatibility endpoint for the callers you can't change. 3. Run both side by side. It costs much less than it sounds, and your migration stops waiting on someone else's roadmap. • Everyone works California time, developers included. • The developer who does the work estimates it, and AI checks the estimate. • The same developers stay from the first meeting through maintenance. #DotNet #LegacyModernization #API
7
If every automated email you send sounds the same, buyers learn to skip all of them. Abandoned cart, post-purchase and win-back are different moments with different jobs. We write them with brand-voice-aware generation per segment and journey stage, so they don't read like one template with the name swapped. Be honest: which automated email from your company would you rewrite first? #EmailMarketing #Ecommerce #AI
1
11
The Guardian reports AI helped unlock Kurt Gödel's notebooks, said to contain hundreds of proved theorems. Most businesses have their own notebooks: scanned purchase orders, years of notes in the ERP accounting system, spec sheets nobody has opened in a decade. Before asking AI to create something new, ask it what your old records already know. What archive in your business would you point AI at first? #AI #Accounting
4
Don't write a separate rulebook for AI agents. Hold them to the one your people follow. Same roles. Same permissions. Same audit trail. Same idempotency on orders. That's how we built Wardenclyffe: the engine doesn't care whether a buyer or merchant brought an AI. Two rulebooks drift apart. One rulebook means every gap you close for humans closes for agents too. #AIAgents #RBAC #Ecommerce
1
6