Kevin Galligan

LangGraph and Firecrawl seem so natural together FireCrawl has so many options available for smart LLM scraping, and structured output. With LangChain’s LangGraph as an orchestration layer you open a world of possibilities for structured output from unstructured web data - Its all you need 🕸️🔥 Combine FireCrawl & LangGraph 👇

17

2 Comments

Here is the perfect example of easy #corewebvitals wins. Yesterday I was asked to check out this site that was failing the Core Web Vitals and it was a fist class example of everything you should avoid. If your HTML code looks something like this you should fix it today. Let me show you how and why: 1. Self host your important scripts. The first one is an easy one. You can see the site is loading resources from googleapis cloudflare cnd, unpgk and bootstrapcdn. That is 4 NEW CONNECTIONS that can take up a considerable amount of time for each new visitor. Now if you consider that all these scripts are parser blocking you can imagine we don't want extra delays this early during page load. The solution: self host these scripts! 2. Defer your scripts. Did I tell you these scripts will lock the parser. I did :-) In general you really want to avoid script blocking the parser. The easiest way to do this is is by deferring your scripts. Simply add 'defer' to the script tag and your scripts are deferred. There are cases where this will lead to errors and you need some extra work but for this site it would have worked just fine. 3. Do not lazy load important scripts On this page the 'all-js' script has a data-src attribute. This data scr is used to late load the script. While there is something to say for this technique (though there are better options!) you should always be really careful with late loading resources. In this case the resource will create a huge layout shift and delay critical functionality like search and menu. The solution: just defer this script an do not over-optimize!! 4. De-prioritize less important scripts Sometimes you do not want scripts like swiper.js to load this early during page load. In this case swiper is not used in the visible mobile viewport. Ther are many way to fix it, the simplest way would be to just place the script lower on the page. A more advacned solutions would be to load it during browser idle or attach it tot the intersection observer to load it just-in-time 5. Simplify your fonts Web fonts should be done with CSS (self host your fonts & preload critical fonts). Preferably as early as possible during page load to minimize layout shift cause by web fonts. Using JavaScript to load fonts will just make your fonts load later with more overhead! 6 Bonus: do not fake your lighthouse score Can you see thie line that reads 'if (navigator.userAgent.indexOf("Chrome-Lighthouse") > -1)'. That will hide everything within this condition from Lighthouse. Lighthouse is a lab test and it is not used to measure the Core Web Vitals (that is what CrUX is for). Lighthouse will try to tell you why your page is slow. You should let it :-)

82

9 Comments

Let's be clear: trust in WordPress and Automattic is broken for good, due to hostile and erratic decisions like this one. At this point, I strongly recommend to avoid Wordpress for the next website you consider building, and turn your attention to other CMS on the market. Remember that security, stability and trust have a cost. https://lnkd.in/e-4BzeDF

5

If Not React, Then What? https://buff.ly/3ODVNAM While I’d like to be spending most of this time working through improvements to web APIs, the majority of time spent with partners goes to remediating performance and accessibility issues caused by “modern” frontend frameworks (React, Angular, etc.) and the culture surrounding them. These issues are most pronounced today in React-based stacks.This is disquieting because React is legacy technology, but it continues to appear in greenfield applications.Source: If Not React, Then What? – Infrequently Noted (https://buff.ly/4fR2vzz)

4

I've seen clients waste millions on overcomplicated software before partnering with MEV. It can be hard to recognize wasted spending when you're in the middle of a process, too. We had a client who was facing this scenario: • A $2 million system, bloated with features • Complex workflows no one needed • Unfriendly user interface What did they really need? Much simpler requirements. The truth hit hard when we stepped in. We stripped it down. Focused on the essentials. Delivered a solution at a fraction of the cost – just $100,000. The result? A system that did exactly what was needed. Quick, efficient, and without the unnecessary frills. The lesson was clear: Saying "no" to good ideas can be tough. But understanding and focusing on what truly matters saves money, effort, and time.

9

1 Comment

Why has authentication become so complicated that there is an entire breed of Saas companies dedicated to it? Supertokens, Auth0, Clerk, Ory, Supabase, FusionAuth, Zitadel are just some of the companies out there offering 3rd party authentication services. "Auth is difficult and risky, don't do it yourself" is the generally accepted wisdom. But then I started looking into Clerk's under the hood implementation, and realized that they regenerate a new JWT token from their front end library every 60 seconds. No wonder they charge per "daily active users". So how did we end up there? 20 years ago you would just have a login, a password, generate a session id, store it somewhere (cache or database), put it in a cookie and be done with it. And a simple password recovery by email. Now, you have tokens, refresh tokens, OAuth 2.0, OICD and it's such a horrible mess. So much of a mess that there are companies that are basically selling a pretty looking form (with username, password, and Google Login button) on a recurring subscription. All of this because of "stateless auth". JWT tokens that can be verified independently of the service that issued them. But really, who needs stateless auth? Only the giants with countless (micro)services need that. Most of us do not. Most likely your service has a database with users anyway for other purposes, so your service is stateful in the first place and dependent on this database. The pain of having to manage and sync users in two different services is much bigger than the problem it solves. So just generate a damn session id, secured in a http only cookie. Store it hashed in your database, and cache it in memory in your application. It takes much less time to implement than researching auth vendors. And it is much easier to make sure it is secure than the tons of services and libraries you end up depending on for such a simple thing!

13

3 Comments

With Velt (YC W22) we provide features that boost engagement and growth within your product. You can add full-stack features like contextual comments, recording, notifications and more. You can fully customize these components. Make them feel native. Make them your own. Match your brand. Developers have full control: * Custom component layouts * Your own CSS * Dynamic template variables * Conditional component rendering * Custom behaviors and functionality * Component variants * Or build your own UI with our APIs. Learn more👇

29

9 Comments

“Codegen tools are broken.” They churn out boilerplate, need endless fixes, and rarely feel… smart. So when I asked Codebuff to create a browser-based Minecraft game, I wasn’t sure what to expect as I just told it: 👉 “generate a basic minecraft clone that runs in my browser that lets me move around in the world, explore it, and mine blocks. keep going until it's very recognizable as a minecraft game.” And here’s what happened: ✅ Hover effects to show where blocks drop. ✅ Particle effects when blocks are destroyed. ✅ An instruction overlay showing block types. Codebuff uses two agents: 👨‍💻 A Developer agent to write features. 🕵️‍♂️ A Product Manager agent to critique and refine. The best part? The game actually works. We showed a basic version to a dozen people, and each added their own ideas and features to our project. Check out the demo video for the final version. Check out our Substack post for more information: https://lnkd.in/gVFp4H69

17

18% of all folks who engaged with the Supabase project, made a contribution somewhere in open source. On top of that, they average about 32 unique open-source contributors weekly. We are standing on some business at OpenSauced, shipping faster than we are announcing publicly (we will fix that announcement cadence soon). In this video, I walk through some new insight features of the project. Let me know if you find it useful. #opensource #insights

62

4 Comments

Can a Hookdeck review get any better than this? "Hookdeck is one of the best software products I've ever used! I look forward to using it for many more years." That same review is followed up by: "IMO, no need for too many new extra features (other than spinning up integrations!), keep it lean and performant" We get absorbed with everything wrong, but customers care about what's right. By focusing on the problems, do we overlook what made users fall in love with the product in the first place?

37

4 Comments

Live from New York, it's billions of rows in the browser 😎 Internal tools should be data-centric, not built as front-ends grafted onto your data. This means: 1. Tools can be auto-generated from data rather than configured in fiddly visual IDEs 2. Performance doesn’t suck when you pass 10k rows 3. You unlock agents that actually know how to operate your tools - not rely on weird browser RPA Had a blast presenting what Howard Zuo and I have been building at Dataland at the inaugural South Park Commons NY Demo Faire. Check our demo here: https://lnkd.in/e5wNrNjA Check the other incredible live demos here: https://lnkd.in/eHWp8U_S. All were great - my favorite was Speaksage :)

80

3 Comments

Version 1.5.3 of our Novashare plugin is live! X broke/deprecated some of its old Twitter code. 🐦 We've pushed out a quick fix where the old Twitter domain link was redirecting and duplicating the query string parameters passed through. If you're using X, we recommend updating. ___ - Added an additional post object check to prevent a PHP warning in certain cases. - Updated our staging site license key exception list with additional formats for Rocket.net (*.onrocket.site) and Pantheon Platform (*.pantheonsite.io). 🔑 - Minor style adjustments to the plugin UI. Changelog: https://lnkd.in/gHuddWW2  #WordPress #smm 

5

If you have a large portfolio of domain names listed on Afternic and want to change your lowercase listings to CamelCase, you're probably going to need a hand to make it happen.

7

Anyone at Apple working specifically on UICollectionView (or maybe generally UIKit?) mind checking FB15040864. To me it seems that something with `group` `count` is broken in Compositional Layout. To reproduce, simply run "Compositional Layout -> Getting Started -> Two-Column Grid" in https://lnkd.in/efNxii7d

1

2 Comments

Version 2.2.7 of our Perfmatters plugin is live! - *New* perfmatters_used_css_below filter. 👨‍💻 - *New* perfmatters_defer_js_exclusions filter. - *New* perfmatters_delay_js_fastclick filter. - Added additional DOMDocument flag to parent exclusion filters for better compatibility. 🔧 - Added GiveWP request parameter to excluded page builders array. - Updated delay JS quick exclusion for WooCommerce single product gallery with additional scripts to help with zoom + lightbox functionality. 🖼️ - Moved Script Manager CSS to stylesheet printed inline instead of from a PHP file. - Updated delay JS quick exclusion for Cookie Notice with additional scripts. 🍪 https://lnkd.in/gy_U_Rk #WordPress #webperf

17

2 Comments

When you quantize down an LLM model, can you maintain its accuracy? I used to think no, but the data says otherwise. For example, Llama 3.1 405B FP16 has an average Accuracy score of 86.63. If you quantize down this model from FP16 to INT4, its Accuracy score falls to 86.47. That’s a recovery ratio of 99.81%. 🔥 🔥 In return, you get 4x more VRAM and higher performance. As people struggle to fit their models onto 8x H100s, a tiny loss in accuracy for 4x more VRAM sounds like a great bargain. Secret sauce seems to be the GPTQ algorithm, combined with neural networks being robust to quantization. Source: Neural Magic's Llama-3.1 70B and 405B INT4 models on Hugging Face. The link's in the comments. #opensource #llama3

35

5 Comments

🤔 What are your thoughts on this controversy? https://lnkd.in/eQ_HHxs2? ▶ On one hand the content was contributed openly and has been publicly available on a third party platform that strives as a commercial entity to be profitable. ▶ But on the other hand many developers contribute this content to voluntarily help the community and understandably some are upset that those contributions are being fed into other tools and platforms to profit. This is a highly complex area that will span across many other User Generated Content (UGC) scenarios were those users will want to weigh in on their rights on content they've published - whether it's just Q&A, comments, articles, audio, video, art, music, recipes, etc.

5

4 Comments