Sarim Malik
@sarimrmalik
Followers
918
Following
12K
Media
270
Statuses
3K
@RubricLabs CEO
Toronto — Islamabad
Joined August 2014
Excited to share more about Rubric. We're an applied AI lab helping companies deploy intelligence. After months of building heads down and working with some amazing companies, I sat down with my co-founders to talk about our process and share more about our approach.
Introducing → Rubric Labs. We're an applied AI lab. We build ambitious AI products. Our team has worked with leading-edge companies like Gumloop and Graphite to help bring practical AI applications to market. We sat down to talk about our process.
4
1
32
Context-free Grammar (CFG) is an incredibly exciting update. This opens the door for products to have their own domain-specific language (DSL) fit for their use case (e.g. data queries, integration orchestration etc.). CFG lets you lock the model into your DSL so it learns the.
This is insane: GPT-5 supports context-free grammars. You can define your DSL and constrain the output to it. The alternative was having some model-compatible JSON representation that was converted into your DSL. So many little things adding up here.
1
0
5
Generate videos in just a few seconds. Try Grok Imagine, free for a limited time.
264
88
1K
In an era of distraction and self‑doubt, trust your gut and build with conviction. Then sprint like there’s no tomorrow. You can will ideas into reality. Momentum is a force vector. You set its launch angle and thrust, the trajectory adjusts as you go.
0
0
7
Tldr: Make it work, make it good, make it fast. Building for simplicity when your product doesn't even work is a recipe for bad UX. "Make it work, make it good, make it fast" is the single best principle to follow when building. You can't make it simple/good when it does not.
complexity first, simplicity second. people say “keep it simple,” but most approach it backwards. they start from simple, then add on complexity without seeing the whole. that’s how you end up with frankenstein products: clean-looking components awkwardly stitched together, held.
1
0
16
RT @MaxMusing: Shipped a new map chart to @Basedash so you can visualize geo data from any data source 🗺️
0
3
0
"If I was to do everything at Stripe again, the thing that I think we could maybe foreseeably and beneficially done differently would be to have spend even more time than we did on APIs and data models.".
A conversation with @patrickc on old programming languages, software at industrial scale, and AI's effect on economics/biology/Patrick's daily life. 00:15 - Why Patrick wrote his first startup in Smalltalk.03:35 - LISP chatbots.06:09 - Good ideas from esoteric programming
0
0
5
My main gripe with the current state of agent frameworks is that they are not grounded in real use cases. Hypothetical abstraction can only get you so far.
0
0
0
It’s clear the industry is transitioning from seat-based to usage-based pricing. Not only is this a better mental model for aligning incentives between vendor and user, but is also a great way to manage your cashflow. You bill the precise cost of supporting a user to the user.
Stripe's usage-based billing platform has grown 145% YTD. There's lots of discussion about when the industry will shift from seat-based pricing to consumption models, but it's clear in our data that the transition is already happening. I'm curious what the second-order effects.
0
0
2
Realization: LLM work belongs in a queue, not a request handler. Submitting a chat message should trigger a background job, the UI streams or notifies when the job finishes. This keeps the UI snappy, avoids request timeouts, and lets you run multiple actions in parallel.
1
0
6
A mistake I made as a first-time founder and see most first-time founders making the same mistake: raising venture capital thinking it's an accomplishment or a rite of passage to building a business. But it's really day 0 of your startup, and definitely not necessary. If.
0
0
3
Shipping velocity is directly proportional to data structure quality, especially in the age where UI can be increasingly treated as an ephemeral asset. More practically, this includes:.- the type of store you use e.g. relational, graph etc. - the way you structure your data.
the more i design/build, the more i realize:. the wrong data structure is like a receding hairline. you're cooked. trying to cover it up makes it even worse . spoke to other founders who agreed that data structure is the ceo's job. every engineer knows to avoid migrations.
12
28
263
Oh man, such a good blog post:
stack.convex.dev
A deep dive on how Convex's reactive database works internally.
2
0
5
RT @arshdilbagi: Most AI products fail in the first month. Not bad AI. Bad prompts. Teams at Discord, McKinsey, Salesforce, DoorDash, Refo….
0
112
0
"It's like there's a certain minimum number of days, weeks, or months (for different tasks) that you need to spend in play mode just collecting signals, accumulating them, and digesting them. And digesting them just takes a certain amount of time, after which you just come out.
Here's something I realized about most of my output. I doubt it's unique, but for some reason I've not heard this concept elsewhere yet. Most of what I produce is not fundamentally bottlenecked on resources (time, energy, etc.) but on intuitive conceptual clarity. It's like.
0
1
9
> Run eval script.> Prompt o3 to plan next feature.> Post on 𝕏.> Catch up on your reading list.> Repeat.
0
0
4
Recently tried a product which is purpose built for data-analysis but in practice found that o4-mini-high in ChatGPT (which is optimized for visual reasoning) just gave me a much better result. This is shocking, because how is a horizontal tool like ChatGPT outperforming a.
0
0
1
Platforms >>> Products. Giving users blocks they can assemble to fit their use case easily and quickly is much higher ROI (and fun) than building for a specific use case. There will clearly be a need for vertically optimized software but thinking horizontally is just a better.
1
0
3