More from Eliran Turgeman
A friend asked me for some study/productivity tips, and I figured the most productive thing I can do is write a post about it. That way, it might help more people too. So here we go. Before we start, there are two things you need before any productivity advice will work: Be introspective. Be honest with yourself. My productivity rules Plan your day the night before Don’t leave decisions for your foggy, maybe lazy, morning self. Here’s the loop: wake up → review study plan → study → write down the plan for tomorrow → sleep → repeat. Check your energy during the day If you just ate and feel sleepy, don’t force deep study. Take a 30–60 minute break - nap, walk, exercise, or scroll your phone a bit, then come back refreshed. You can also take micro-breaks: finish a chapter, grab water or a piece of chocolate, and get back to it within five minutes. Don’t lie to yourself about effort Setting goals is great, but be honest about how hard you actually worked. You can check all the boxes, feel proud, and still know deep down that you took it easy. That’s fine sometimes, we all need rest days - but don’t confuse that with an intense study session. Review your day When you plan tomorrow’s tasks, reflect on today. Did you actually do what you set out to do? If not, why? Maybe your goal was too ambitious, or maybe you just spent too much time gaming. Either way, learn from it. There’s always room to improve, if the goal really matters to you. Limit distractions Put your phone on silent and out of reach. If you study on your computer, close anything that might tempt you, and even hide shortcuts to distracting apps. When I was in uni, I played a ton of League of Legends. The desktop icon was staring at me every time I opened my laptop — so I buried it under three folders. Sounds dumb, but it worked. A few more things Get good at breaking big goals into small tasks. If your goal is to pass an exam, start by mapping out all the smaller steps that’ll get you there. Spread them out over time, with a bit of buffer. Plans will change — that’s fine — but you should always know where you stand and adjust as you go. And one more time, because it’s that important: don’t lie to yourself. If you spent four hours on TikTok, felt bad, then studied a little to compensate — that’s not a “productive” day. Call it what it is. It’s okay to have those days, you’re human — just plan for them instead of pretending they didn’t happen. I believe that much of the productivity advice online includes some stupid ceremonies and whatnot, just do what you feel works for you. I think the ability to introspect, being honest with yourself and improving over time is all that matters. Take whatever breaks you need, in whatever order you want. You don’t need a fancy journal to write your tasks, open a txt file. You don’t need a perfect system, just something simple that works for you.
Most people solve their subscription fatigue by canceling Netflix. I solved it by vibe-coding my own workout app instead of paying another SaaS. Two weeks ago, I decided to get serious about my workouts again and start logging them. I looked for an existing solution that has the following: create workouts log sets, reps, and weights a calendar to track consistency simple enough. Free apps were crammed with ads. Paid apps had bloated features I didn’t want. Both annoyed me. Before vibe-coding, I’d either tolerate ads or pay. Now there’s a third option - build my own. Of course, you could always build your own, but pre-vibe-coding it would take much more time to be worth it. How did I choose the vibe coding platform? I logged into loveable, base44, bolt, and wrote the following (imperfect) prompt 1 2 3 4 5 6 7 I want to create a personal webapp for managing my own workout routines (kind of a workout logger) I want to be able to define "workouts" - collection of exercises including sets and reps I want to be able to track which workout i did on which day - calendar view I want to be able to log the weights I did for every exercise in every set. I want my workout templates to be a simple collection of exercises that are plaintext - don't create some kind of an exercise library.. i just want to type the exercise name myself make it stateful, including a db connection to store all the relevant data. Whichever tool gave me the best first shot, I ran with. This time it was Loveable. Iterating With that one-shot starting point from loveable, I published it, and went to my first workout at the gym, all excited and ready to use what I built. First workout: I needed notes for exercises. Another: supersets. Each time I wrote it down, went home, and 15 minutes later I published a new version with loveable. After two weeks of using this app at the gym and doing tweaks, the app feels solid. On my last iteration, I added a badges/achievement page, and a github-like consistency widget that looks cool I hope will help me stay on track and be consistent. Sharing Friends wanted to try it too - the problem? I didn’t make it secured by sign-in, I thought I only need it for myself, and even if someone’s going to find this weird loveable URL, they could only see my workouts - who cares… But of course for my friends I’ll write one more prompt - and so I added Supabase auth within minutes. Final Thoughts This post isn’t about showing off my little workout logger anyone could make in a few hours prompting. It’s about how easy it is today to scratch your own itch. You build the exact features you need, when you need them. No ads, no bloat, no adapting to someone else’s UX. Just something comfortable and fun to use. It’s never been easier to bring your ideas to life, small or big.
If there’s one pattern I’ve seen across multiple companies, from scrappy startups to big corps, that causes endless headaches, it’s this: a single cache cluster shared across services. I recently shortly wrote about my lessons from building and maintaining distributed systems at scale, and the first point that came to mind is exactly this - it starts with an excuse of simplicity, “we already have a cache cluster up and running, let’s just make this other service use it, no need for more infra”, and ends with a confused on-call engineer trying to debug which services were affected by the last keys eviction. So I want to double down on this idea and explain in more detail why it becomes a nightmare once your system scales. One eviction policy You got different services each throwing keys at the same redis cluster. A sudden spike/bug just caused a dramatic increase in cache writes - your cluster wasn’t ready for this, it hits maxmemory and now different keys are being removed based on your eviction policy. What’s the problem? there’s no isolation - service A caused the max memory, and now service B, C, D also pay the price - their keys are being removed as well from the cluster, and could affect the latency, and correctness of other flows of your system. Monitoring is harder Our metric fires up — we see a drop in hit rate on the cluster. Which service is causing it? Who’s affected? Instead of thinking about one service, you’re now mentally juggling everything across the entire system. More noise, less clarity. Although monitoring is harder, you could set up application monitors that you send once you write/read from the cache, based on the prefix of the key. potentially if you are organized and each service that uses the cluster has a unique prefix and you can easily identify between the hit rates of different prefixes - that’s great, but you have to work to get there. Debugging is harder This ties back to my first point about the eviction policy. You had 10m keys. something happend. now you got 5m. The effect on the services is really hard to trace. One service might have lost 100k keys, and you barely see a difference in its monitors, but it doesn’t mean your users are not feeling something is off, maybe today the are waiting a bit more for the page to load, but it’s not too long to hit your monitors thresholds. In that case, if you didn’t have a monitor on the cache cluster for keys eviction, you might be totally blind..”oh I see a slight latency increase here, but no monitors popped - guess all is well” So, never use a shared cache cluster? No, that’s not the lesson here. In some cases it is totally fine to use a single cache cluster. For example: You don’t really have a lot of traffic read/written to the cache so most of it is free anyway You store shared static data (for example, feature flags) Also note that some of the points I was making here against using a single cache cluster, can be somewhat mitigated by having good monitoring set in place. For example, having a defined prefix for the cache key per use-case, per service, and publishing metrics in the application level so we have observability to which type of keys (by prefix) are experiecning a low hit ratio. But on the other hand, tracking keys eviction is harder to monitor, since it’s not initiated by your system. Anyway, I hope you get the point. If you are getting started, a single cache cluster is totally fine. Otherwise, spin up another cache cluster, and sleep better at night. 🚨 Become a better software engineer. practice building real systems, get code reviews, and mentorship from senior engineers. Get started with 404skill
I recently wrote about over-engineering and striking a good balance between making your code “too” future-proof and not making it future-proof at all. Some time later, I realized it was missing a critical perspective. I hadn’t addressed over-engineering from an architectural point of view, so this post is dedicated precisely to that. Let’s talk about a decision I made for Collecto, my side project. Collecto is still in its early stages, and like most early-stage projects, its future is uncertain. It could grow into something big—or not. That’s where architectural decisions get tricky. You don’t want to overengineer and waste time, but you also don’t want to under-engineer and regret not laying a solid foundation. So what’s the problem? Collecto is a forms-backend service, meaning it handles the creation, management, and processing of forms data for applications. I wanted to add the ability to send emails on certain events. For example, when a new user signs up for your form, you might want to send them a welcome email. The simplest solution? I could write a new service responsible for sending emails and call it directly wherever needed— for example, right after a user signup is saved to the database. This approach works, is easy to set up, and introduces no additional overhead. However, it results in tight coupling, making future changes more challenging. If tomorrow I want to also send a notification to the form owner when they receive a new subscription, I would have to keep adding more responsibilities to the form service code. This bloats the core service, which should ideally focus solely on CRUD operations for forms. On the other end of the spectrum, I could go all-in and build a distributed pub/sub system with a service bus like RabbitMQ or Azure Service Bus. This would give me scalability, decoupling, and all the good stuff. But it’s also a massive investment in time and complexity for a project that doesn’t need it, yet. I didn’t like both options, so I looked for a 3rd alternative and found MediatR which is a mediator pattern implementation in .NET. Why MediatR is a good middle-ground? MediatR facilitates communication between different parts of the application without them needing to reference each other directly. Instead of invoking methods directly, you can send requests or publish notifications, allowing registered handlers to respond accordingly. This approach maintains loose coupling, making the system easier to maintain and evolve. At the same time, it doesn’t introduce the overhead of managing infrastructure like a service bus or message queue. Everything stays in-process, simple, and fast. One of the primary reasons I chose MediatR is its simplicity. Implementing communication patterns with MediatR is straightforward and requires minimal configuration. Compared to a full-fledged service bus, MediatR demands a much smaller time investment and eliminates operational overhead such as monitoring queues or scaling message brokers. It can’t be all sunshines and rainbows MediatR has a few cons compared to other out-of-process messaging brokers, for example Events are in-process only. If your application crashes, you lose the events. There’s no out of the box retry mechanism for failed event handlers. If you deploy multiple instances of Collecto, MediatR won’t distribute events across them. Bottom line Architecture isn’t about perfection—it’s about trade-offs. MediatR worked for Collecto because it gave me a decoupled, flexible way to handle events without the overhead of a service bus. It wasn’t the simplest solution, but it was the right one for where the project is today. The next time you’re making an architectural decision, remember this: the best solution isn’t the most impressive or complex—it’s the one that solves your problem now while leaving room for growth later. 🚨 Become a better software engineer. practice building real systems, get code reviews, and mentorship from senior engineers. Get started with 404skill
More in programming
Helping Scout focus and remember
tl;dr: the Explorer Pattern is one individual or team building a functional implementation that minimizes cross-team and cross-functional impacts as much as humanly possible, and then reconciling the details with other teams after that prototype is completed, accepting that this approach can require very substantial rework to that initial development. Why write this? Imprint previously called this the “Tech-led” or “Engineering-led” development model, but it’s pretty unclear to folks what that means. This writeup provides a more defined term, the “Explorer Pattern”, and explains in more detail how this works. Why do we need to adjust the way we work? Before 2026: Code is expensive, and we design process to minimize implementation and rework In and after 2026: Writing software is extremely cheap, but it’s a wider phenomenon: generating output is effectively-free across many dimensions. Code is nearly free — if we have appropriate constraints. Designs are nearly free — if they follow patterns. Analysis is nearly free — if we’ve instrumented things properly Reworking software is extremely cheap. Historically industry has optimized around avoiding engineering rework because engineering was the highest cost / narrowest constraint. In many companies this was interpreted as writing code being the constraint, which was never true, but many common practices were designed as if that were true (e.g. bringing engineering into problem-solving late “to protect their time”) Working on the system is much more effective than historically if we get automated integration and feedback loops right, e.g. not improving a one-off design but improving the design system. Any place where improving the higher-leverage process is slower than solving the one-off is a bug in our process Explorer Pattern: primary-and-replicas metaphor The easiest metaphor to think about the Explorer Pattern is database replication. Databases like MySQL and PostgreSQL have a primary database, where writes are performed. Every write is verified at that point, and once it’s verified, you know that the writes will flow downstream to various secondary replicas. Those downstream replicas can be delayed but they can’t be wrong because writes are verified upstream. (I’ve resisted the desire to talk about the CAP theorem.) A different model for database design is that used by Dynamo or Cassandra (or MongoDB to a certain extent), where you have a replica set, and have to get an acknowledgement from the majority of the replica set for a write to be strongly consistent. This is a very reliable setup, because it means you can tolerate losing a node, but it also introduces inconsistencies, and on a fixed amount of hardware they tend to be much, much slower than running a primary-secondary setup (or rely on bad performance metrics that don’t fsync to disk to create the illusion of throughput). The Explorer Pattern is the former primary-secondary model, where the vast majority of the work is done by the working individual or team. This makes all work internally consistent because it’s passed through the same small group. It doesn’t, however, mean that the work is perfect. For example, you can always insert data that meets the table’s requirements but is still inherently wrong (e.g. 200 is a valid integer, but it doesn’t mean that my credit limit was supposed to be 200, that depends on the surrounding application logic). Other teams become aware of the changes downstream, not synchronously upfront, and it’s instead the working group’s responsibility to both rationalize approach and create visibility into their work for other teams to engage in a non-blocking fashion. Example: Support for Financial Aggregators A concrete example of this practice at Imprint was rolling out financial aggregator support, e.g. allowing folks to connect to their Imprint powered accounts via Plaid. This has always been something folks want, but it’s particularly important in the realm of agentic commerce where more folks’ agents are analyzing their financial state for them. For this project, one of our engineering leaders (Ron), built the first pass of the feature at a low-burn while understanding the shape of the problem in a private alpha. He chatted through some of the details periodically to keep other teams aware, but it was a bit of a side project. However, once he found the full shape of the problem and had implemented a full solution, we cut the replica-lag across teams down to near-zero, identified and ran through the small number of remaining issues, and shipped it. Like many of our projects this year, it was never officially on a roadmap, and shipped a quality feature that our members have wanted for some time. How to follow the Explorer pattern The rough rules for following the Explorer pattern: Know your goal and sponsor: every initiative ultimately has to have a specific outcome it’s pursuing and a sponsor who’s bankrolling that pursuit. This doesn’t always need to be explicit, especially for smaller projects, but once a project goes beyond ~3 weeks of effort, if you can’t very explicitly identify who the stakeholder is sponsoring your project and they would not enthusiastically agree, then you are likely closer to “lost” than “exploring” Small working group is small: at most 5 people, and preferably smaller. Everyone in the group has to keep the entire project in their head for as long as possible to detect problems and opportunities Small working group is flexible: folks in small working groups have to “find a way to yes” rather than “find a way to no”. Exploring is about uncovering the fundamental shape of a problem rather than pretending we know the fundamental shape Publish progress in a public channel: explorers have to report home with the progress they are making so others are aware. This is both a Linear Project and a Slack channel. Make the narrowest viable changes to what already exists: when we introduce unnecessary changes, we invalidate others’ mental models about how the world works. When we haven’t made those changes with care, we’re not only invalidating others’ mental model, we are creating the expectation that we’ve done it very intentionally for a good reason, which has a chilling effect on others improving them later, which is very expensive, whereas the Explorer pattern depends on being as inexpensive as possible Exercise good judgment to identify reasonable hypotheses: discard bad ideas immediately, and then filter the rest by trying to implement them as quickly as possible. Once you see the shape of the implementation, you can figure out which of the remainders are interesting Hold assumptions constant to avoid getting stuck: the only way to navigate complex ambiguous problems is to hold things constant, even when you know they are imperfect. For example, to do headcount planning we have to know our revenue, to know our revenue we have to forecast it, to forecast it we have to know the day that partners will go live next year. We don’t truly know any of those things, so just anchoring on a plausible guess to unblock us is much more valuable than trying to get them accurate. A little prework is useful, but a lot is just artificial precision that’s a distraction from progress Hold space for being totally wrong: when we move fast, with low information, we always find the way out, but often it’s not the way out we anticipated when we got started. Being open to throwing it all away, and starting a second, third, fourth time Ship all the way to production behind a feature flag: features and products move fast when we get the pieces working every step from frontend to backend to data pulls. This is the first step of exploration. Just make sure it’s behind a feature flag that’s only enabled for internal folks!
Comparing two different ways of work and working environments.
I own an old, cheap TV. It has no Internet connection, no fancy smart features that make you want to install a Pi-hole, and no ask-me-again-later updates. However, it is connected to my own desktop computer through an HDMI cable, so I can watch movies and TV series from the couch. Setting it up used […] The post Making a dumb TV just smart enough appeared first on Quentin Santos.
Five days and 540km on a bike across Kyushu, without a laptop. Notes on the peace of riding all day, convenience store dinners, and Japan's emptying countryside.