About
Founding Engineer at Quesma, working on token economics, efficient agentic coding, data…
Activity
2K followers
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted thisExcited to share that our paper MaSC has been accepted at NeurIPS 2026 (Evaluations & Datasets Track)! 🚀 Coming shortly after my first-author acceptance at SIGGRAPH Asia 2026, wrapping up the first year of my PhD with two A* venues is an incredible milestone. In this work, we tackle a major challenge in personalized text-to-image models: evaluating concept preservation reliably without relying on expensive, non-deterministic LLM-as-a-judge APIs. MaSC (Masked Similarity Metric) leverages subject masking to evaluate both concept retention and prompt adherence in a single forward pass with a frozen SigLIP2 encoder: - Matches GPT-4o agreement with human ratings on DreamBench++. - Outperforms prior open metrics (including DreamSim and DINOv3) while staying open-source, deterministic, and API-free. - 0.992 AUC on ORIDa real-world object recognition across environments. 🌐 Project page: https://lnkd.in/dNZpJURK Huge thanks to my advisor Wojciech Palubicki and co-authors Lennart Petersen, Bartosz Kotrys, Dominik L. Michels, and Soren Pirk for the great collaboration! Looking forward to Sydney and Kuala Lumpur this December. #NeurIPS2026 #SIGGRAPHAsia2026 #GenerativeAI #ComputerVision #MachineLearningMaSC: A Masked Similarity Metric for Evaluating Concept-Driven GenerationMaSC: A Masked Similarity Metric for Evaluating Concept-Driven Generation
-
Bartosz Kotrys shared thisMalgorzata Piotrowska, Cezary Piwowarczyk, and I had a great time at the first Tokenomicon in Amsterdam, organized by the Tokenomics Foundation and FinOps Foundation. We wrote a recap of our favorite talks, practical lessons on AI costs, and where coding-agent billing still needs work. One takeaway: fewer tokens don’t always mean a smaller bill. You can read it here 👇 https://lnkd.in/dUst3Biz It was great to finally meet and talk with Stephen Arthur and Kevin Emamy from the FinOps and Tokenomics Foundations in person. I really enjoyed the conference format, and Henrique Amorim deserves a special mention for the energy he brought to the stage. A real talent! Thanks also to Łukasz Kubicki, Rafael S., ☁️ Jean Latiere ☁️, Ali Rahi, Yuriy Prykhodko, Rense Siegmund, Andrea Blasioli, Jan C. Simons, Paweł Faszyna, Sebastian Amrogowicz and Magdalena Rapacz-Matras for the great conversations. Hope to see you all again soon!Token ecomonics in Amsterdam: Inside the first Tokenomicon - Quesma BlogToken ecomonics in Amsterdam: Inside the first Tokenomicon - Quesma Blog
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted thisThis week at Quesma: Malgorzata Piotrowska, Cezary Piwowarczyk, and Bartosz Kotrys attended the sold-out Tokenomicon + FinOps X in Amsterdam, alongside 500+ practitioners who own AI budgets. Our technical benchmarks are gaining traction. Jacek Migdał spent two days with peer founders at the Inovo.vc Magia Summit. Piotr Migdał taught the next generation of engineers at the Model Trainers workshop. VCs are knocking too. Piotr Bukanski from Acurio Ventures visited our Warsaw office. Meanwhile, we merged many PRs and dogfooded our product. Next month, we will launch our product: analytics for agentic coding and go on a US tour: SF Bay Area, Oct 19 to 30, with a booth at AGNTCon + MCPCon in San Jose on Oct 22 and 23.
-
Bartosz Kotrys reacted on thisBartosz Kotrys reacted on this🎉 MaSC has been accepted at NeurIPS 2026 (Evaluations & Datasets Track)! Personalized text-to-image models are judged on two things: 🔹 Concept Preservation: is the generated subject still your dog, your backpack, your sneaker? 🔹 Prompt Following: does the scene match the text prompt? The standard metrics (CLIP-I, DINO, CLIP-T) score the whole image for both. People don't judge it that way: we compare the subject with the reference and check the background against the prompt. MaSC builds such reasoning into the metric. With a mask of the subject, one forward pass of a frozen SigLIP2 encoder computes both scores. 📊 Results: ✅ Concept preservation on DreamBench++: MaSC reaches Krippendorff α = 0.506 against human ratings, just ahead of GPT-4o (0.499). It beats GPT-4V (0.426) and every open metric we tested, including DreamSim (0.420) and DINOv3 (0.343). As far as we know, it's the first open, non-LLM metric to match an LLM judge on this benchmark. ✅ ORIDa real photos: AUC = 0.992 at recognizing the same object across different environments. ✅ Prompt Following: beats CLIP-T at no extra cost. For years, open metrics for concept preservation were stuck around α ≈ 0.3. MaSC reaches LLM-judge level while staying open-source, deterministic and API-free. 📦 Try it: pip install masc-metric Huge thanks to my co-authors Patryk Bartkowiak, Lennart Petersen, Bartosz Kotrys, Dominik L. Michels and Soren Pirk, and to our reviewers and AC. See you in Sydney 👋 📄 https://lnkd.in/d_PSmQjw 💻 https://lnkd.in/dvVFqsSY #NeurIPS2026 #GenerativeAI #TextToImage #ComputerVision #Evaluation
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted this🎉 MaSC has been accepted at NeurIPS 2026 (Evaluations & Datasets Track)! Personalized text-to-image models are judged on two things: 🔹 Concept Preservation: is the generated subject still your dog, your backpack, your sneaker? 🔹 Prompt Following: does the scene match the text prompt? The standard metrics (CLIP-I, DINO, CLIP-T) score the whole image for both. People don't judge it that way: we compare the subject with the reference and check the background against the prompt. MaSC builds such reasoning into the metric. With a mask of the subject, one forward pass of a frozen SigLIP2 encoder computes both scores. 📊 Results: ✅ Concept preservation on DreamBench++: MaSC reaches Krippendorff α = 0.506 against human ratings, just ahead of GPT-4o (0.499). It beats GPT-4V (0.426) and every open metric we tested, including DreamSim (0.420) and DINOv3 (0.343). As far as we know, it's the first open, non-LLM metric to match an LLM judge on this benchmark. ✅ ORIDa real photos: AUC = 0.992 at recognizing the same object across different environments. ✅ Prompt Following: beats CLIP-T at no extra cost. For years, open metrics for concept preservation were stuck around α ≈ 0.3. MaSC reaches LLM-judge level while staying open-source, deterministic and API-free. 📦 Try it: pip install masc-metric Huge thanks to my co-authors Patryk Bartkowiak, Lennart Petersen, Bartosz Kotrys, Dominik L. Michels and Soren Pirk, and to our reviewers and AC. See you in Sydney 👋 📄 https://lnkd.in/d_PSmQjw 💻 https://lnkd.in/dvVFqsSY #NeurIPS2026 #GenerativeAI #TextToImage #ComputerVision #Evaluation
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted thisStudents had one Friday evening, 18 GB of coding-agent transcripts, and a single question: where are we burning tokens? They investigated wasted tokens, agents stuck in loops, and whether swearing at AI changes its behavior. This was Quesma’s first token economics hackathon. Please read our blog post. Link in the comments. Congratulations to Ekipa (Paweł Juzek, Natalia Sołoducha, Maja Kowalska) for winning, with a special mention to Zespół 1 (Michał Sarzała, Piotr Z., Tomasz Święcki) for my favorite presentation. Thanks to Kevin Czupryński (STARTUP FOUNDERS STARS), Bartosz Podgórski, Tomasz Swieboda (Inovo.vc), and Malgorzata Piotrowska (Quesma) for organizing.
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted thisQuesma is proud to sponsor #AGNTCon + #MCPCon North America, October 22–23 in San Jose, California! 🚀 Deep technical sessions, real-world case studies, and two days of agent infrastructure in one room. Come find us to talk #tokenomics! 🎟️Register now + join us 👉 https://bit.ly/44MI7LM 🔥Explore the schedule: https://bit.ly/4f6Pp35
-
Bartosz Kotrys reposted thisBartosz Kotrys reposted thisLast week we open-sourced Quesma Shipper. It mattered more to us than it sounds, and I’ll explain why. As I hinted a couple of weeks back, we’re building a tool that shows companies where their coding agent spend actually goes. To achieve that, we need a process running on developer machines that, in the background, reads agent sessions. We collect prompts, tool calls, all transcripts. It’s a lot to ask of anyone so we decided the best way to show that you can trust us is to show the code. quesma-shipper is now public, licensed under Apache 2.0, written in Go, and able to run on Mac, Linux and Windows. We strip all the secrets on the developer’s machine and encrypt the payload. Data can be uploaded straight into a bucket you own. Repo link in the comments. Still early but go ahead and poke at it a little bit.
-
Bartosz Kotrys shared thisLove to see `awesome-ai-tokenomics` listed in recent AI Economics Weekly Newsletter. Thanks Victor Garcia, FinOps Weekly and AIEconomics Weekly for recognition. Links in comments:
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisI've joined MEGA.dev - this time it was not a choice, but destiny. A four-week program on rethinking how we build, with Theo Browne, Angie Jones, Kent C. Dodds, and John Lindquist. Starts October 26. - mega.dev People are deterministic, same are AI Agents. Change my mind. My participation in mega.dev was no different - I got my ticket from a fun lottery. Thanks Grzegorz Róg for organizing a fun game where I've learned more about my personal agentic workflows. It already pinpointed for me: what are my agentic workflow habits, and where I can improve. More details in the comments! It's time! #BRAVE #ItsTime #Megadev
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisExcited to share that our paper MaSC has been accepted at NeurIPS 2026 (Evaluations & Datasets Track)! 🚀 Coming shortly after my first-author acceptance at SIGGRAPH Asia 2026, wrapping up the first year of my PhD with two A* venues is an incredible milestone. In this work, we tackle a major challenge in personalized text-to-image models: evaluating concept preservation reliably without relying on expensive, non-deterministic LLM-as-a-judge APIs. MaSC (Masked Similarity Metric) leverages subject masking to evaluate both concept retention and prompt adherence in a single forward pass with a frozen SigLIP2 encoder: - Matches GPT-4o agreement with human ratings on DreamBench++. - Outperforms prior open metrics (including DreamSim and DINOv3) while staying open-source, deterministic, and API-free. - 0.992 AUC on ORIDa real-world object recognition across environments. 🌐 Project page: https://lnkd.in/dNZpJURK Huge thanks to my advisor Wojciech Palubicki and co-authors Lennart Petersen, Bartosz Kotrys, Dominik L. Michels, and Soren Pirk for the great collaboration! Looking forward to Sydney and Kuala Lumpur this December. #NeurIPS2026 #SIGGRAPHAsia2026 #GenerativeAI #ComputerVision #MachineLearningMaSC: A Masked Similarity Metric for Evaluating Concept-Driven GenerationMaSC: A Masked Similarity Metric for Evaluating Concept-Driven Generation
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisGreat to be at #FinOpsX (aka #tokenomicon) in the beautiful city of Amsterdam this week. With 2,000 attendees, including practitioners and a record number of vendors, the energy around this emerging FinOps challenge was obvious… The dreaded #AI cost had blossomed like the famous tulips this region is so known for. Hopefully its not an omen of things to come though... Many speakers (almost all of them) had AI as their focal discussion point, but one thing I did notice throughout was the vast differences we heard about the benefit of #PromptCaching and optimisation at different points of the whole prompt process. Numbers are clearly still settlingso the industry is clearly still calibrating these metrics and best practices. Remember, you can see our approach in #Informa to this rising challenge here: https://lnkd.in/e7attEBF And while I try to digest the numbers, and the differences between them, I would like to say that I am a fan of two things that surfaced in the event. 1. Production, consumption, and value are three distinct focal points and perspectives that need different capabilities, observations, and services. FinOps has been looking at consumption mostly up until this point, but many leading experts have rightly predicted that we need to move out of ‘cost per token’ (consumption) and into ‘cost per outcome’ (value). Repeating the car metaphore again, we need to not only measure the price per litre (token), but also the mile-per-gallon (result). 2. Splitting activity out of the basic observations and into obvious sections of engagement like helping select the right engine to service the token, picking the right level of precision, routing/caching, and governance are all touch points we need to be involved in. Final thought. Consumption (or production) of tokens, gateways, and building systems that utilise them, all need clear oversight for cost, usage, quality, and safety, so the parallels for this are eerily similar to the early challenges we had with Cloud costs. I feel we’re back at the beginning of the same-but-different-this-time challenge that gave rise to #FinOps for Cloud.
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisLook mom! I made it to the big screen. My earlier message on LinkedIn made it to the keynote in Amsterdam. Thanks Henrique Campos Amorim for the shout-out. Glad you found your Power Ranger color 💙. It’s morphing time! ☝️Dips on green by the way!
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisBist du immer noch 🐝b.. Izzzing von FinOps X Amsterdam? Ich weiß, ich bin es! Alte Freunde treffen, endlich Leute persönlich treffen, neue Kontakte knüpfen und irgendwie ein weiteres großartiges Gespräch führen, das an jeder Ecke auf mich wartet. FinOps X fühlte sich wirklich wie ein summender 🐝Bienenstock neugieriger und sachkundiger Menschen an – sie tauschten Ideen aus, forderten sich gegenseitig heraus und waren wirklich begeistert davon, wohin FinOps sich entwickelt. Und da KI scheinbar Teil jedes zweiten Gesprächs war, gab es viel zu besprechen. Außerdem konnte ich an der Agentic AI in FinOps: From Insight to Autonomous Action Chalk Talk bei Tokenomicon teilnehmen, die an KI-Anwendungsfällen arbeitet und über das Eigentum an agentischer KI in Unternehmen diskutiert. (Dean Oliver hat mir sogar ein Mikrofon 🎤😄 gegeben – erster Fehler des Tages :D) Spaß beiseite, ich liebte das Format: weniger Präsentation vor Leuten, mehr das Zusammenbringen der Community, um zu diskutieren, zu hinterfragen und zu verarbeiten, was agentische KI für FinOps tatsächlich bedeuten könnte. Ich kam mit vielen Ideen, vielen Bildern und vor allem vielen großartigen Gesprächen nach Hause. An Erik Marke, Marc Hesse, Jan C. Simons, Jens Gößler, Daniel Silva, Andrée Lundberg, Peter Isemann, Rob Martin, Ben de Mora, Jaspal Kaur Sandhu, Ruben Vander Stockt , Laura-Vanessa Soldner, Jannick Berghäuser, Max Zug, Christopher Klein Laura Wieser und alle, die ich getroffen, getroffen, von dem ich gelernt habe oder mit denen ich gelacht habe: Danke. ❤️ Ich freue mich schon auf das nächste. #FinOps #FinOpsX #Tokenomicon #AIFinOps #AgenticAI
-
Bartosz Kotrys liked thisBartosz Kotrys liked this2 days Amsterdam Tokenomicon and FinOps ended today and it was great being back on track and met quite a view team mates from the past adding new ones to my network. There was a lot of great content and session to follow from a keynote to lightning talks and breakouts a lot to digest. And the AI train left the station through the "golden gate" ( sorry to much a MAC guy) to value creation and if not probably not even get a seat in second class. Highlighting the session from Robertas Vasaris around how Pharma with their scientific to AI approach and open the FinOps Das the Portfolio Manager of Intelligence. Loved the host Henrique Campos Amorim guiding us through the session, still have to learn to pronounce his name right and looking forward for a more interactive role inside FinOps and Tokenomicon Foundation in the future. Or in other words think I am back in the game combining my new way of contract negotiations in the AI Era with the stronghold of the Linux Foundation bringing FinOps, ITAM and AI together resonates a lot. Looking for more and it was great to meet and talk with Juha Kaukoniemi, Rich Gibbons, Christian Bosshard, Jeremy Chaplin, Thomas van der Mark and more folks who are already decades in the business.
-
Bartosz Kotrys liked thisBartosz Kotrys liked thisFirst of all - wow, Tokenomicon FinOps X Amsterdam was an incredible event. Thanks to J.R. Storment for putting the trust in me to help facilitate the Tokenomics Foundation member working groups and then asking me to talk about that work - on the big stage! That was the product of many weeks of group effort, and I'm so happy that it's all live now on https://lnkd.in/g3MFBZ9A The analysis of the first State of Tokenomics survey has also just been released! I'm thrilled to be able to point to the top findings and challenges you reported having and have the opportunity to talk about solutions. There is a ton more work to do as we dig into the results - together, and in the open. https://lnkd.in/gRjw6d_K It was also so cool to finally meet (some of) my teammates for the first time as coworkers in Amsterdam - Kevin Emamy Steve Trask Jessica Brandes Henrique Amorim Ishita Vyas Dean Oliver James Newton Sam Benson Sara Whitehead Joe Eletto Rob Martin AJ Witt I had a number of deep and insightful conversations with faces old and new, and I love the energy you're all bringing to Tokenomics. My top takeaway is the validation that we're moving in the right direction: actively prioritizing and working on the biggest problems you all face. Let's keep the conversation going - please continue to share questions and ideas with me here or in Slack!
Experience
Education
Courses
-
Angular 4 - DevMeeting (2017)
-
-
AngularJS - DevMeeting (2014)
-
-
Business Negotiations - persuasive techniques (2019)
-
-
Deep Dive into Javascript with Kyle Simpson (2018)
-
-
Docker & Kubernetes - Bottega (2019)
-
-
Problem Solving Team Leader - Lynsky Solutions (2019)
-
-
React and MobX - DevMeeting (2017)
-
-
Recruitment and selection of employees (2019)
-
-
Responsive Web Design - DevMeeting (2015)
-
-
Rethinking Async with Kyle Simpson (2018)
-
-
Web Application Security Training - attacking and protecting web applications (2019)
-
Languages
-
English
Full professional proficiency
-
German
Elementary proficiency
-
Polish
Native or bilingual proficiency
-
Spanish
Elementary proficiency
Recommendations received
8 people have recommended Bartosz
Join now to viewOther similar profiles
Explore more posts
-
Clean Canvas
🚀 Save 100+ Dev Hours & Boost Your Margins with Clean Canvas 🚀 What if your next Shopify build didn’t come with hidden costs… only boost? At Clean Canvas, we don’t just deliver themes, we engineer performance, clarity and scalability directly into the code. That means no tangled Liquid files, no CSS conflicts, no endless debugging… just clean, modular code that works. 💥 Why it matters: Modular architecture and strict component isolation let your developers locate and fix issues fast, no more hunting through monolithic files. Built-in features (like filtering, mega-menus, quick view, and more) replace common third-party apps, saving hours of setup, integration and maintenance. Optimised performance (minimal scripts, lazy loading, organised CSS) ensures faster build-outs and better user experience, straight out of the box. 💥 The result? Your agency saves over 100+ dev hours per standard build, significantly improving project margins and leaving time for what matters... innovation. If you’re ready to stop swapping time for shortcuts and start delivering premium Shopify stores that convert and scale, we’ve got your back. Curious how clean code saves agencies 100+ dev hours? Read the full breakdown 👉🏼 https://lnkd.in/e3uAB2-H Or browse our themes 👉🏼 https://lnkd.in/e7Xr8r8z #Shopify #Ecommerce #WebDevelopment #CleanCode #AgencyGrowth #CleanCanvas
-
Jared Wray
Just released a new #javascript / #typescript library called Hashery. 🎉 It does object hashing with some nice features: - Browser and Nodejs Compatible as it uses the web crypto API - Supports SHA-256, SHA-384, SHA-512 (WebCrypto), plus DJB2, FNV1, Murmer, and CRC32 - Easily replace JSON parse and stringify with custom functions Check it out: https://lnkd.in/dnb9kFHR
-
Nicos Nicolaou
🚀 KSP 2.3.4 has been released! The Kotlin Symbol Processing (KSP) team just shipped a new update focused on performance, stability, and cleanup. What’s changed in 2.3.4: ✅ KSP now ignores sources located in the Kotlin directory ⚡ Avoids recording Java symbol lookups in non-incremental builds, improving efficiency 🧹 Proper cleanup of ThreadLocals once processing is complete 🔗 Full changelog (2.3.3 → 2.3.4): https://lnkd.in/gdbx4KhH #Kotlin #KSP #AndroidDevelopment #BuildTools #DeveloperTools #OpenSource
-
Tom Hemming
🚀 CodeLeft Version 2 is here! What’s new? We’ve introduced a Bring Your Own Model (BYOM) architecture for code assessment. Here’s what that means for you: 🔒 Privacy first – your code never leaves your IDE. 🛠️ Full control – you decide who sees it and where it’s processed. 🤖 Smarter reviews – pair our new Community Edition of CodeLeft with GitHub Copilot for AI-powered code reviews. 📚 Skill boost – get insights on SOLID principles, Clean Code, and Complexity analysis. 🧹 Tackle tech debt – upgrade your codebase while levelling up your own coding practices. 👉 Dive into our new architecture and explore all the product updates here: https://codeleft.ai/
-
Adebayo Bamijoko
How to add a working contact form to a serverless portfolio in under 5 minutes. No backend. No API route. No excuses. I recently did a small revamp of my portfolio website. After struggling for almost three years with “should this be a founder profile or a software engineer portfolio?”, trying several variations, I finally found the right blend. While working on the About / Contact section, something hit me. A lot of portfolio websites have contact forms that don’t work. Or worse, they’re just UI. Placeholder inputs. No submission logic. And honestly, I get why. Most portfolios are deployed on serverless platforms like Vercel, Netlify, or GitHub Pages. No server processing. So people default to email links, LinkedIn buttons, or WhatsApp. But here’s a simple workaround. I use Google Forms as the backend. – Create a Google Form – Get the full form action URL (not the short link) – Extract the entry.xxxxx field IDs – Submit via fetch using FormData – Done Works perfectly on serverless deployments. You’ll see a code snapshot attached showing exactly how the submission works. Downside? Google Forms doesn’t always notify you instantly. My workaround is simple: I check responses every few days. If you’re building a portfolio, a client site, or a simple company page on a serverless setup, this saves you time and removes friction. A contact form should work. Not just look good. Hope this helps someone. ________ 👋 Hi, I'm Adebayo Bamijoko. I’m a Software Engineer focused on architectural simplicity and efficiency. This year, I’m sharing the "senior" shortcuts and patterns that save hundreds of hours of coding. Hit connect so you don't miss a post.
-
Bernard Lim
Most code is built on lies. TDD tells the truth. At WunderWaffen, we help big companies use AI to work better. One thing we talk about is TDD, or Test-Driven Development. Think of it like building the Marina Bay Sands in Singapore. You calculate the structural load before you pour the concrete. 🏗️ In TDD, you decide how straight the walls must be before you even touch a brick. In 2003, Kent Beck wrote the definitive book on this. He showed that writing a "test" before writing the "code" saves a lot of money. It stops mistakes before they happen. The logic flows like silver rain, to wash away the coder's pain. 🌧️ Now, AI models like Claude 3.5 Sonnet are making this even faster. AI can write the tests for us using frameworks like Pytest. It uses "mocking" to pretend it is testing a real system. This means we can find bugs in seconds, not hours. Global tech leaders rely on this discipline to prevent outages like the recent CrowdStrike incident. Reliability is not an accident; it is engineered. As we move toward a world where AI does most of the work, TDD is the key. It is the map for the machine. Without a map, the AI will get lost. In silent rows the numbers stand, shaped by a digital, ghostly hand. 👻 If you are a CEO or a business owner, you should care about this. Good tests mean your software won't break when you need it most. It means your team spends less time fixing errors and more time making money. Are you using AI to test your code yet? Comment CLAUDE if you want to read more and I will send it to your inbox. Make sure you connect first! p.s. A bug found today saves a fortune tomorrow. 🪿 Comment your thoughts below. 🥧 Share this to spread the word. 🧨 Follow for daily updates on the future of tech and AI. #TestDrivenDevelopment,#AIAutomation,#SoftwareEngineering,#DevOps,#WunderWaffen,#FutureOfWork,#TechTrends,#CodeQuality,#Agile,#MachineLearning
-
Denis P.
🚀 New feature dropped: Statement of Work Launched DeployProof yesterday. Got immediate feedback from 2 potential customers: "We need scope approval BEFORE starting work, not just screenshots after." So I built it. Now on Pro plan you can: 📋 Define deliverables, timelines, payment terms ✉️ Send to client for approval (no login needed) ✅ Client approves/rejects each item ✍️ Electronic signature 📄 PDF evidence with SHA-256 hash The full cycle: 1. SOW approval → start work with clear scope 2. Screenshots → prove what was delivered 3. Client signs off → legal proof Stop scope creep before it starts. 👉 https://lnkd.in/dBiuQ8C6 #SaaS #Freelancing #WebDevelopment #BuildInPublic #Bootstrapped
2 Comments -
Matt Mochalkin
Stop exposing your Doctrine Entities to GraphQL. 🛑 If you are building Enterprise GraphQL systems with Symfony, tightly coupling your API to your database is the #1 architectural mistake I see. It kills flexibility and leaks implementation details. In my article, I break down how to architect a fully decoupled, type-safe system using Symfony 7.4 and PHP 8.4. We dive deep into: ✅ The Decoupling Rule: Why you need DTOs (Data Transfer Objects) and how to map them efficiently. ✅ Solving N+1: Practical strategies using Data Loaders and eager loading. ✅ CQRS Mutations: Keeping your processors thin by dispatching Messenger messages. ✅ Modern PHP: Leveraging PHP 8.4 Property Hooks for cleaner logic. This isn't just a "hello world" tutorial. It’s a blueprint for scalable logistics systems. Read the full guide here: 👉 [https://lnkd.in/dXmmVBsJ] 👉 [https://lnkd.in/d_3Y-hQp] #Symfony #PHP #GraphQL #SoftwareArchitecture #APIPlatform #WebDevelopment #EnterpriseSoftware
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Bartosz Kotrys
3 others named Bartosz Kotrys are on LinkedIn
See others named Bartosz Kotrys