Tiago Loureiro

If you are looking to contribute to a good, active, and deep-tech OSS project, you will find DiceDB pretty neat. We have opened many interesting, low-hanging good-first issues. These issues are a great starting point to understand how production-grade databases are actually built. The codebase is still in its nascent stage and it is relatively easier to get started with it. So, go for it :) github.com/dicedb/dice by the way, DiceDB is the world's first truly real-time database with SQL-based reactivity. We are breaking down Redis and architecturally changing it to become optimized for modern hardware - making it multi-threaded - added SQL support - native JSON support, and so on If you cannot contribute patches, you can help with the reach by leaving a star on the repo and sharing this post :) #AsliEngineering #DiceDB

766

11 Kommentare

We benchmarked Apollo vs WunderGraph vs Mesh vs Grafbase GraphQL Federation gateways. Here are the results (and my take on the benchmarking): CONTEXT: Last year, we launched Grafbase GraphQL Federation (built in Rust) from scratch. To see how it performs compared to others, we’ve started benchmarking it against other providers. Here’s the full benchmarking setup: - All gateways had the same CPU and RAM resources, limited by Docker. - We repeatedly executed a complex query using the load-testing tool K6, with a constant and ramping VU scenario. - The final response content is 88KiB big, merged from several subgraph responses. Now to the interesting part… RESULTS: Throughput: - Grafbase Gateway: 1502 req/sec - Cosmo Router: 867 req/sec - Apollo Router: 434 req/sec - Mesh: 300 req/sec Latency: - Grafbase Gateway: 57.9 ms - Apollo Router: 92.4 ms - Mesh: 127.2 ms - Cosmo Router: 308.7 ms (We’ve benchmarked more than just throughput & latency, the full blog post is linked in the comments below) MY TAKE: First, I’m happy to see that all the hard work we put into the Grafbase Gateway paid off. I believe it’s mainly due to two things: 1/ We betted on Rust to build the Grafbase Gateway right from the start – and that’s still paying dividends today. Some other core design choices also contributed to this result. 2/ As a company, we’re striving for a culture of performance. This goes beyond our engineering team, but definitely reflects in product performance. But what does this result mean from a business perspective? 1/ First, great performance helps you reduce your infrastructure footprint. For enterprise companies, we’re talking about billions of requests per minute (yes, minute). These seemingly small differences quickly turn into hundreds of thousands of dollars in infrastructure costs. 2/ It’s a better foundation for scaling. This is especially relevant for large enterprises. If you expect heavy load on your systems (because you want to scale), your foundation needs to be rocksolid. At this level, there’s no room for small performance issues. 3/ Better user experience Above all, better performance means a better experience for your users. I think I don’t need to elaborate on this. The effects of great (or not so great) user experience can be felt in *every* aspect of your company. – To be honest, I was also impressed by some of the results. The sustained latency vs. throughput metric showed that you can keep throwing requests at the Grafbase Gateway and get guaranteed great latency. If you’re interested in the full benchmark, check out the link in the comments. All that’s left to say: Kudos to our amazing engineers at Grafbase. – PS - We tried to be objective. If you feel like we should benchmark other metrics or see ways to improve - comment below, file a complaint or open up a PR. PPS - If you encounter a case where Grafbase doesn’t provide this level of performance, we’d be happy to look into it for you.

40

8 Kommentare

The market for third-party CDN services, which I define as the delivery of video and small/large objects, was about $5 billion in 2023. (Full blog post: https://lnkd.in/ekFZqX9y) This figure does not include revenue for services tied to security (WAF, DDoS, etc.), compute, hardware, and storage and excludes delivery in China. The number excludes any company with its own DIY CDN, like Netflix and YouTube, that doesn’t sell its infrastructure as a service to other companies. The figure also excludes revenue from vendors offering online video platforms, including Brightcove, Uplynk, AWS Media Services, Quickplay, Deltatre, Mediakind, Sports Radar, Cognizant, Endeavor Streaming, StreamAMG, Bedrock Streaming, etc., which combined saw revenue of about $1.8 billion in 2023. Based on what I know, AWS Media Services is the largest of the group, with about $450 million in revenue in 2023. While reports produced by research firms say the CDN market was multiple times larger than my number, they are wrong. Full stop. Do not buy their reports. Akamai, Fastly, Edgio, CDN77, and others disclose enough details tied to their delivery revenue that we have numbers to form a good foundation for market sizing. The market is not as big as many suggest, and those writing about the size of the CDN space don’t define it and include non-CDN services like transcoding, hardware, hosting, storage and transit in their numbers. Full blog post: https://lnkd.in/ekFZqX9y Listen to my special CDN-focused podcast from July 2024 for more details and updates regarding 4K, low/ultra-low latency, CMCD, DIY (TikTok, Disney), Open Caching, pricing and multi-CDN. https://lnkd.in/epzXSMVA Vendors mentioned: Akamai Technologies, Amazon Web Services (AWS), Edgio, Fastly, Cloudflare, CDN77, Medianova CDN, MainStreaming, Qwilt, Wowza, Ateme, Broadpeak, Gcore, Netskrt Systems Inc., Varnish Software, Vecima Networks Inc., Velocix.

666

39 Kommentare

Currently, rollups are fragmented, leading to poor UX and disjointed liquidity. Here the first demo of Javelin by our portco NodeKit With Javelin, users can synchronously send messages and move funds across #rollups on NodeKit within a single 2s block.

11

Model Finetuned on MonsterAPI beats LLaMa 3 8B!! We hosted a LLaMa 3 Hackathon 2 days after the model was launched, the team at MonsterAPI worked day and night to get the infrastructure ready for the participants. Despite the small scale of the event, developers from countries like France, the United States, South Korea, the United Kingdom, Kenya, and others participated in the Hackathon. The 3 days of the hackathon were a challenge for the participants as their goal was to beat the LLaMa 3 8B and surpass the baseline score on these benchmarks: • MMLU (Massive Multitask Language Understanding) • Winograd Schema Challenge • GSM8k (Grade School Math 8k Problems) Devs worked hard, fine-tuning the model countless times on several data sets to try to achieve the goal. Finally, on day 3, Eswar Divi was able to surpass the benchmark scores and take home the Xbox Series S. Eswar excelled with a Winogrande score of 76.56% and a GSM8K score of 69.83%, securing the top spot by fine-tuning LLaMa 3 at the cost of just $5, showing the potential for further developments with larger fine-tuning experiments. These results showcase the power and importance of fine-tuning and how it significantly significantly boosts efficiency and accuracy. For those who didn’t take a prize home, keep your spirits high and watch for more future hackathons on our Discord channel - https://lnkd.in/dC4Dhcv8. The hackathon was a blast for us, the team, and the participants. Join our discord channel to get early access to updates about the upcoming hackathon, or just take part in engaging conversations about the world of generative AI. Team MonsterAPI

17

The conventional advice for building horizontal products is to choose a single vertical and expand one by one 👀 Yet within the first 6 months of launching our SDK generator, we built support for 7 languages: Python, TypeScript, Terraform, C#, Ruby, Java, and PHP. With only 4 FT engineers at the time, was this the right decision for the business? Here’s what we learned: *️⃣ Building so many language targets in a short period of time really pressure tested our generation capabilities. We built a platform that was flexible and easy to develop on because we HAD to. *️⃣ We figured out how to access expertise from language communities that we were non-native to. For example, we’re not native PHP developers, so we’re working with people like 🤓 Steve McDougall who know the languages inside & out to keep us on the right path. *️⃣ Perhaps most importantly, we listened to our customers. Companies in different industries need different languages. C# is very important in Fintech. Most people are not interested in a Terraform provider. But for anyone whose API is considered infrastructure by their users, Terraform is critical. Focusing on breadth over depth drove early momentum for us and accelerated us into customer conversations. Each language took a good amount of time from a small team. But rapid experimentation helped us quickly validate which industries had specific language demands as well as which industries needed a bit more education on the value of SDKs. Was every language we built worth it? Head to our platform to start generating SDKs and decide for yourself! Link in the comments 👇 #sdks #langauges #api #pmf

16

3 Kommentare

We've spent quite a lot of time improving Tapir's documentation lately: 1. we've got a new landing page - written with people yet unfamiliar with Tapir's and Scala's ecosystem in mind: https://lnkd.in/dkRT_kPT 2. there's a listing of how-to's, in the form of examples; each example is tagged, described, self-contained and easily runnable using scala-cli. Just copy the content of a single file! https://lnkd.in/dyJknFwx 3. finally, we've got a growing list of tutorials, each in a textual and video format: https://lnkd.in/eUKniQkT That way, we aim to address all 3 aspects of good technical documentation: tutorials, how-to's and reference documentation (that's what we mostly had so far). By no means "Tapir docs" is a done project, but we certainly got better foundations now! #Scala

70

There is a $7.5B company that: • Started with $0 funding • Has 100M+ active users • Generates $1B+ in ARR And now powering the open-source revolution. Here’s the story of GitHub: Tom and Chris were both working at CNET when they met in the mid-2000s. They were both passionate about software development and knew that developers needed better tools to manage and share code. In 2008, they started working on a side project that would change their lives. Git was a version control system at the time. But it had terrible UI and no functionality for collaboration. Their idea: build a web-based platform to host the Git repository. It would revolutionize the way developers collaborated on code. They launched GitHub in April 2008 with just a handful of users. Early on, they bootstrapped with an intense focus on customer experience and on exceptional UI. And they fully embraced the open-source community. The commitment paid off... Three years later, they had over a million developers on their platform. GitHub became the focal point of the open-source revolution. Hosting and collaborating on code was never so easy. Even big tech (FB, MSFT, Google etc) became customers. From there things skyrocketed 👇 By 2012, the biggest VCs started noticing. GitHub raised $100 million in Series A funding led by Andreessen Horowitz Their growth was explosive. By 2015, they had 9 million users and hosted over 21 million repositories. How did they scale so rapidly with a relatively small team? Simple answer: They leveraged talent all over the world. GitHub adopted a distributed team model early on, hiring talented developers from around the world. Scott Chacon, a co-founder, worked remotely from Mexico, setting the precedent for a global workforce. GitHub attracted top talent from regions like Eastern Europe etc. GitHub's distributed model also allowed for continuous development cycles with teams in different time zones. By 2018, GitHub had revolutionized the developer experience. And Microsoft bought them for $7.5 BILLION. All the co-founders are billionaires. Not bad for a side project. GitHub, WhatsApp, Wayfair and many giants were built on global talent. Trained pros take on rote execution at a fraction of the cost, while top teams focus on strategy and product. GrowthAssistant embeds elite, college-educated pros from the Philippines directly on your team. We find and place talent with the EXACT skillset you need. DoorDash, HubSpot, True Classic, and 200+ others use us Check it out:

23

2 Kommentare

Boom! We just hit over $2M USD in TVL on Lofty's staking portal for the first time! This is under 1 year since we launched this feature and there are over 50 pools left to deploy. And we're just getting started. Two things we've learned from this that's been surprising: 1. We thought that this feature would almost be exclusively used by crypto native people, while our real estate users would completely ignore this. Oddly, it's been the opposite. The vast majority of funds come from real estate users who already buy/sell the properties. Most of them aren't crypto natives and it's probably the first time in their lives that they've used a DeFi product, especially staking. Even when we chat with potential sellers, it's one of the features that they're most interested in learning about + trying out. 2. We thought more people would be interested in staking the real estate tokens, because you can double dip in lending yield + rental income. Under the right conditions, you'd make more money than just investing in real estate. But it turns out most people prefer to stake USDC. This all goes to show that you can sometimes be directionally correct, launching this feature would be a massive boon to our business and that instant liquidity is truly something that real estate investors/owners want, but be wrong about the specifics like who the target users should be and what they would like the most about the feature. Never a dull day, when you're building.

57

8 Kommentare

How often does your static analyzer struggle to identify the source code nuances? It probably happens more often than you'd like, doesn't it? In the article, our team shares how we've dealt with it. Let us introduce you to our new user annotation mechanism! #pvsstudio

1

With the announcement today that Momento's Topics has gone GA, wanted to bring this content back up to the front. If you are looking for rock solid consistency, amazing performance, and the option to bring your own language to connect to their SDK, you need to be checking it out. Here's an article I wrote about Topics and Webhooks with #rust and their SDK https://lnkd.in/ghHuR-cC

13

1 Kommentar

🤖 Creating a fully autonomous AI developer to write and maintain code is still the holy grail of software development. There are many brilliant teams currently working on this, and each of them have their own unique approach. The team at Cosine have made a massive step towards making this a reality. “Cosine achieves state-of-the-art results on the SWE-bench benchmark” They have achieved a score of 43.8% on SWE Bench, which was verified by OpenAI. What's even more incredible is they have managed to do this with a lean team and a relatively limited budget. Though it’s important to note Cosine isn’t looking to replace human developers entirely, but rather augmented them with human-like assistants that they can collaborate with on any kind of coding task. Yang Li + Sam Stenner + Alistair Pullen Read more about their collaboration with OpenAI 👇 https://lnkd.in/gu7xXVPn

21

More #rust and #serverless this week. Momento is working hard on an SDK for Rust and I've been putting a few sections of it through its paces. It's no secret I love the Momento software but I'm more impressed by the company's talent, internal culture, and their commitment to customers. In this article, I explore using their Cache Topics to process messages. But I do that as a comparison between using their Subscriptions via the SDK for Rust AND with the Webhooks product and a Lambda Function with a Function URL. Both of the modes stand up well, perform amazingly but come with different profiles for different use cases. I cannot recommend them strongly enough and I'd be happy to put you in touch with anyone over there if you are curious. It truly is serverless at its finest. Enjoy! #serverless #aws #momento #believeinserverless #rust #performance https://lnkd.in/ghHuR-cC

23

the "ai future" requires 100% open-source. not for code reasons, but for TRUST reasons. push-back is typically "impact to financial potential", so here - https://lnkd.in/gQxk6gfj is an intrinsic trillion $ to get things started on that debate or defend the goal of being less trustworthy ;> @eff @opensourceorg #linux #code #trust #ai #postaitrust

2

Whatever this shit, engineers need to show some backbone and refuse to commute 40 miles in bumper to bumper traffic congestion to build it for him. Particularly applies to engineers in Bay Area. Computing paradigms are meaningless when the need is for a new transportation paradigm that addresses problems like congestion and safety. I for one is quite nauseated with this over innovation in computing when the human needs are elsewhere. How does that matter to me if Andrej Karpathy’e LLMOS deals with tokens or bytes - it’s not my interest to aid him in his mental masturbation. We need a new operating system for the roads (particularly highways) to improve operational efficiency, relieve congestion, improve safety and user experience. Operating System needs to get back to its fundamentals. How about moving people instead of bytes or tokens with efficiency and basic dignity if not user experience? The context window is a platoon of “real” heterogenous vehicles including buses and cars moving at highspeed insync on a highway.

2

1 Kommentar

Bowen Yang from Cohere will dive into the world of long-context models and how we can push the boundaries of transformer architectures. At TMLS2024 we'll explore: ✅ Scaling Transformers: Techniques for handling longer inputs while maintaining efficiency. ✅ Challenges & Solutions: Examining the modelling and framework hurdles we face, along with potential solutions. ✅ A Range of Approaches: We'll delve into methods across data, modelling, and framework levels. ✅ Training, Evaluation, & Inference: Covering the entire NLP pipeline for long context tasks. This panel promises to be a valuable resource for anyone interested in pushing the limits of language understanding. See the full abstract here: https://lnkd.in/gwSRjPjz #NLP #Transformers #LongContext #MachineLearning Toronto Machine Learning Society (TMLS)

29

You do realize that 2GB of ram means 2B bytes? For 3B params? And you need multiple bytes for each token? You can quantize all you want - LLMs will never be reliable in limited memory save for narrow purpose-build cases, that’s just physics. (Some would say they will never be reliable, period. )

3

3 Kommentare