March 20th, 2026
Over the past couple of weeks our team has been split into 3 major efforts.
- Bringing the production line up to speed for the remainder of MP1, MP2 and preparing for MP3.
- Implement new features and improvements to the firmware
- Data cleaning and structuring for the CRM
Production
Production Line Improvements
The new production floor has allowed us to greatly improve our workflow and effectiveness. We knew that our improvements would help reduce errors and improve efficiency but now we have actual data to support it. The cosmetic rework ratio on the Main Line has been reduced by ~66.4%! This was one of the biggest slowdowns so this improvement will translate directly to making more machines faster and with better quality.
MP1
Production restarted, it took longer than we anticipated. Switching to the new production line was not as smooth as we would’ve hopped. But the good news is that the production line is running again and we re-started shipments. ~80 machines have either shipped or are waiting for pickups and we are on track to ship the remainder of the 1000 units in the next few days.
MP2
Material readiness is still not 100% solved for MP2. We improved our processes a lot since MP1 but we still need to adapt in real time to use our time as efficiently as possible with the materials we have on hand. This is not that surprising. We are still “early” into mass production and the program is not mature. Both our suppliers and our CM are still adapting to the many different things that can go wrong.
Our biggest risk continues to be the funnel. We are working on several solutions in parallel and we feel confident that this will not create additional delays but we still don’t have parts on hand.
Just like with MP1, we’ve been able to make a lot of progress by focusing on subassemblies, if we receive the funnels in the next couple of days like we expect, we should be still on track to finish at least 50% of MP2 by March 31st. The remainder 50% could take as little as 1.5 week but we won’t know until we get more information this week. We will keep you posted.
Software
Having several hundreds of users making coffee every day has been amazing.
- Bugs are now found way faster and are way easier to diagnose/fix. During development we would sometimes hunt a bug for days or weeks. Now, any minor problem is reported by several people simultaneously with enough details that makes the process way easier.
- We are getting a “pulse reading” of what people care about the most. Some of these are not necessarily bugs but are significant improvements that make the experience and the coffee better!
1. Improved control during “low flow” periods (Available in Nightly)
We overhauled our control algorithm to increase stability at very low flow rates (~0.5ml/s for example).
At higher flow rates, the previous method was ok. But it started to become super noisy around 1.5ml/s and it was really noisy at under 0.5. The new method improves this dramatically.
2. Improved pressure-flow controller switching (Available in Nightly)
When switching from flow control to pressure control or viceversa, the system now handles the transition with a lot more grace!
3. Ignoring cup adjustments and minor bumps (Available soon in Nightly)
Stop-by-weight is awesome but because our scale is very sensitive, adjusting the position of the cup or touching the scale at all during brewing would bring the weight over the target momentarily, causing the shot to stop prematurely.
The new algorithm is a lot smarter about weight and is able to ignore minor bumps and cup adjustments during brewing. It is still the best practice to not touch the cup while brewing to get the most accurate results, but at least this should prevent shots from being ruined by a last second adjustment. This is how it looks!
REMEMBER:
Nightly is the most unstable version. Only use it if you want to help us test new features and find bugs. Otherwise, wait for these changes to arrive to Stable!
Data cleaning
We’ve been promising the CRM for a long time. We have a version that is working pretty well internally. The only problem is that “pretty well” is good enough for internal use, but not for public release.
It should not take much longer, this is something we are actively working on and we want to use to coordinate deliveries for MP2.
