Mistakes
The main mistakes I made across these projects. Each entry records why the error arose and what I learned from it.
Clura Enclosure
Ran too many projects simultaneously, which slowed all of them down
Why it arose: Everything looked worth doing at once, so nothing got a clean run at being finished.
What I learned: Focus on one thing at a time, keep track, and finish it quickly. Speed is everything and should be prioritized over cost. We picked the STM32 over the ATMega because it was cheaper, which led to far more development work writing libraries for every single component we wanted to install.
Never checked whether there was product-market fit
Why it arose: It started as a passion project, so the question of who needed it and why never got asked properly.
What I learned: If I ever aspire to sell a product, even one that starts as a passion project, I should really consider whether there is product-market fit.
Kept adding features while developing the product
Why it arose: Every improvement seemed obviously worth the small delay it cost.
What I learned: This leads to a never-ending cycle of improvement and development which delays launch.
Shared what we were doing far too late
Why it arose: We did organic marketing and tried to share what we were building, but only started right before launch.
What I learned: Sharing along the way is a method to keep yourself accountable, and to see whether people are even interested.
Lost track of time, tasks and overlapping deadlines
Why it arose: There was never a moment where the whole project was laid out end to end, so dependencies only appeared once they were blocking.
What I learned: Before any next project, take at least a day to understand everything it could entail: what is needed at each step, what has to happen in series and what can happen in parallel. Then move forward hitting small milestones.
Built a hardware product from outside China
Why it arose: University keeps me in Europe, and I assumed remote iteration would be good enough.
What I learned: For hardware products you should not stay outside of China. Everything there is fast, efficient and cheap, optimized for prototyping at speeds that are just impossible in the West. Being there could easily have shaved half the time off this project, if not more.
Delegated things I should have done myself
Why it arose: Delegation is a lever that lets you do more, so I reached for it even where the work was on the critical path.
What I learned: Sometimes, if you want things done on time and done well, you should just do them yourself, even if you don't know how yet. Learn, fail, and improve until you can.
Let my emotions get intertwined with the project
Why it arose: When the business is your life, every delay reads as a personal failure.
What I learned: It is normal to get emotional when things go badly, but it leads to poor decisions and an aversion to working on the project, because we usually don't like to experience pain. Detach yourself, or at least keep some section of your life separate from what you're doing. That stability is what lets you partition problems and think clearly. Read the related post →
Clurasense
Didn't understand the actual timelines our customers were working to
Why it arose: We believed October was already too late to start, since the deadline to build the power plants was June 2026. That turned out to be completely false: very few companies will finish by then, and the deadline has since been extended.
What I learned: Fear of being late made us go fully OEM, which raised our costs by at least 30% and made us less competitive on price than if we had developed some of the hardware ourselves. Check the deadline your customers actually operate to before you let it dictate your architecture.
Didn't check European trademarks before naming the company and the products
Why it arose: The naming was treated as a branding exercise, not a legal one.
What I learned: We had to rebrand mid-marketing. Luckily this was caught quite early, which let us land some of our biggest clients under the new name. Check copyrights and trademarks before the name goes on anything.
Overestimated how many replies we would get from direct outreach
Why it arose: We focused on contacting clients and reaching the entrepreneurs who had built the power plants, people who weren't yet looking for anything.
What I learned: 80% of the customers we secured came through Google Ads and our site: people who were already searching for weather stations. Sell to demand that already exists before trying to create it.
Spent an excessive amount on Facebook ads for leads that never converted
Why it arose: The leads were genuinely tied to real agrivoltaic plants, so the targeting looked right, but they were not searching for a solution, so they were not ready to buy.
What I learned: Intent matters more than fit. Put the budget where people are already looking.
Used AWS by default instead of running it on a server we built
Why it arose: AWS was the obvious choice without anyone sizing the actual workload.
What I learned: The AWS bill doesn't really go below $100 a month, and our workload will never be intensive: we store hundreds of bytes per station. Running our own server would have kept the burn lower.
Tried to import SIM cards into China
Why it arose: It seemed simplest to have the SIMs installed at the factory with everything else.
What I learned: Importing SIM cards into China is almost impossible: they are an easy way to violate the firewall. Time and money went into getting a few in. The right answer was to ship everything to our warehouse first, install them locally, and then ship the finished products to customers.
Didn't keep track of every lead in an organized way
Why it arose: Leads arrived faster than the process for handling them.
What I learned: Keep every lead tracked, with a structured and automated way to reach them by email and WhatsApp, and then call them.
2 MW Agri-Photovoltaic Plants
Underestimated the problem of getting the neighbours to agree to the grid connection
Why it arose: The connection route crosses land that is not yours, and I treated that as something to settle later, by which point the other side knows how badly you need it.
What I learned: Get a signed preliminary agreement far in advance, before it becomes the critical path.
Underestimated the impact of the slopes on the field
Why it arose: I relied on a surveyor instead of measuring the terrain myself, so the real gradients only showed up once posts were going into the ground.
What I learned: Get a millimetre-accurate topographic map of the field yourself, with a LiDAR drone. Stake it out in house rather than trusting a surveyor, and run a string between two pillars to check heights before driving a single post.
Underestimated what it takes to unload the material
Why it arose: Containers were shipped before anyone had decided where and how they would be emptied.
What I learned: Unloading heavy goods from a container without a crane is close to impossible: you need either side-loading trailers or a crane, both difficult and expensive. Ideally use a warehouse at ground level and your own forklift, and decide this before the goods ship.
Underestimated shipping times and demurrage at the port
Why it arose: Nobody was ready to collect the containers the day they landed.
What I learned: Every day a container stays in port costs you money. Make sure that as soon as they arrive they are collected and unloaded where they are needed.
Got the order of works and of purchase orders wrong
Why it arose: I ordered things in the sequence I happened to think of them, not in the sequence their lead times demanded.
What I learned: The transformer cabin determines every other timeline, so order it and have it ready as early as possible, and pour its concrete slab immediately, since it takes two weeks to dry. Order the PV cables and the thicker connection cables from China at the same time.
Bought the whole CCTV chain from a single supplier
Why it arose: Site security was treated as one line item at the end instead of a system with its own requirements.
What I learned: Have China ship pre-wired cabinets (batteries, cameras, and a solar street light) together with the structure, and mount them on a concrete block. Watch the channel count on the video recorder: two channels for every thermal camera. Don't rely on one manufacturer for everything: order cameras, power supplies, switches and fibre separately, using pre-terminated armoured fibre. Prefer solar-powered cameras where possible, with wireless radio bridges back to the cabin.
Left the fencing and the site boundaries until late
Why it arose: The fence looked like the last step of the build rather than the first line of protection for everything already delivered.
What I learned: Order the fencing from China, ideally with the robot to install it, or just order the material there and have a local contractor put it up. Check the boundaries carefully, and get surveillance in place before the works start, not after.
Underestimated the landscape and regulatory procedures
Why it arose: I planned the plant as an engineering problem and treated the paperwork as a formality that would sort itself out.
What I learned: Budget real time for it: Opengenio for the seismic analysis, SINPOL/SIP for the preliminary notice of works, updated every time a new firm enters the site or you get fined, then DILA/PAS, landscape constraints, and tree-felling permits. Ask every contractor for its INAIL and INPS details the moment you award the contract.
Didn't check the CE certificates of the critical systems early enough
Why it arose: Certificates were something to collect at the end, after price and quantity had been agreed.
What I learned: We got lucky: everything we installed did turn out to have its certificate in order. That was luck, not process. For every critical system installed, the CE certificate is the first question to ask after quantity and specs, not the last.
Started the works in winter
Why it arose: The schedule was built around order lead times and never around the weather.
What I learned: Rain stops a construction site. Don't begin the works in winter.
Andromeda 3D Printer
Left the software to the very end and tried to achieve alone what a team of 10-15 people did
Why it arose: Inexperience, and lacking someone to work with: I was keeping everything a secret.
What I learned: Don't work alone. Find people who are great at what you aren't. Share your progress and get feedback from people as soon as you can.
Was tied to long iteration times with nothing to do in between
Why it arose: Not structuring, from the beginning, everything that had to be done.
What I learned: Keep track of what you want to achieve, mark progress, identify mistakes. Never wander for months without going back to the bigger picture of it all.
Over-engineered it, which meant testing too many features at once
Why it arose: Being just one more feature away from being done.
What I learned: Build something, ship it. Test it and then continue. Never build for a long period of time without testing.
Picked the project on passion alone, with no thought for the financial side
Why it arose: It was pure enthusiasm for delta printers. I never asked who would pay for it or why.
What I learned: Be disciplined about choosing the projects to work on. Without a monetary incentive it's hard to get people on board to work with you, and making progress will be difficult.
Corsari Energy
Qualified the leads too late in the funnel
Why it arose: We optimized for lead volume, and we were too optimistic about the minimum roof size that would make a project feasible. In the real world none of the roofs we found had the perfect pitch and conditions for that size.
What I learned: Qualify as early as possible, even if it means fewer leads. It is better to spend all of your time on 2 good leads than on 20 medium-quality ones.
Let unqualified people take up our time
Why it arose: There was nothing forcing a prospect to prove their situation before we engaged with them, and many lied just to see what was possible for them.
What I learned: People will waste your time if they are allowed to. Screen as fast as possible with real documents that prove their actual situation.
Advertised to roofs in general instead of a niche
Why it arose: Broad ads were the fastest thing to launch, so that is what we launched.
What I learned: Focus on a niche instead of general roofs. Instead of simple ads, reach out directly to sites you have already identified as suitable: cheaper, and probably more effective, because you already know they qualify.
Automated Hydroponic Greenhouse
Started something I did not have the time to finish
Why it arose: The grant was there and the idea was good, so I took it on while Clura and Clurasense were both still running.
What I learned: An opportunity you cannot give real time to is not an opportunity. Spend your time on what really matters, and let the rest go before it becomes a half-built thing you are still carrying.