AI amplifies ambiguity.
If you don’t know what you’re building, you’ll just get to the wrong answer faster.
IMO this makes thorough planning even more important. Specs and designs force the hard questions early while also giving the agent something concrete to build toward.
Warfa Jibril retweeted
Breaking: Folarin Balogun will be available to play in USA's Round of 16 match against Belgium on Monday, FIFA announced.
The FIFA Disciplinary Committee has suspended the red card issued to the USA striker during their Round of 32 win over Bosnia and Herzegovina.
Better products should save you money rather than accelerate your spend. LLMs allow builders to be lazy because they’re generally decent at most things but we shouldn’t throw out the tools that got us here. Tasteful determinism will separate the great from the good
How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching.
Better Defaults (not Usage Caps) – Engineers can choose any model they want, but defaults matter. We’re experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task. 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models, so they can check each other's work.
Better Routing – In our custom harnesses, we preprocess prompts and route to the best model for the job, considering cache hits and model pricing. For instance, you may want a frontier model for planning, but not for execution where they can be overkill. Ultimately, humans shouldn't be choosing models - AI can automate this task.
Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests are cache aware, so we’re reusing a warm cache wherever possible. For example, our cache hit rate went from 5% → 60% in LibreChat once properly implemented.
Keep Context Lean – Start fresh sessions when switching tasks. Scope file context narrowly. Disconnect unused tools. Don't just compact. The goal isn't fewer tokens used, it's fewer tokens wasted.
Better Visibility – Our engineers can use as many tokens as they want, from whatever model they want, but we’ve made usage visible – and the more you spend on AI, the more impact we expect.
The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable.
Putting this into practice has cut our AI spend nearly in half, while our token usage continues to grow.
Warfa Jibril retweeted
The limit of writing code as AI improves = code review. As developers have less visibility into their codebases, good testing paradigms will become imperative
Warfa Jibril retweeted
As code generation improves, interfaces into models will become more important.
Text is just the beginning 🖌️🎨🖼️
Ideas exist across many levels of granularity. We should be able to express and collaborate on them with humans and AI at every level
Warfa Jibril retweeted
We just fixed the worst part about Browserbase.
Our browser recordings sucked, and it was all my fault.
The team rebuilt it from the ground up, and now it's one of our best features.
It's live today, free for every customer on every plan!
Warfa Jibril retweeted
Today, we’re releasing Power Retention, a new architecture beyond Transformers.
It enables LLMs to handle millions of tokens efficiently, unlocking long-context applications that were too costly before.
manifestai.com/articles/rele…
Insane progress from @pk_iv and @browserbase! Excited to see what comes in 2025 🅱️
Today, we've been named the #1 early-stage company in the Enterprise Tech 30!
But we're just getting started and need amazing people like you to join our team as we continue to scale.
Come 🅱️uild with us.
Warfa Jibril retweeted
Anyone – of any race, creed or nationality – who came to America and worked like hell to contribute to this country will forever have my respect.
America is the land of freedom and opportunity. Fight with every fiber of your being to keep it that way!
🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸
Warfa Jibril retweeted
we’re building a culture of maniacs on a mission. erik (below) is one of them. want to join? check out the open roles on our website.
I recently joined a rapidly growing startup that raised a Series A of $21,000,000.
I joined 5 weeks ago and so far my experience has been a short of amazing. The founder, @pk_iv , likes to shoot for the fences, nah scratch that he likes to shoot for the stratosphere.
My position was essentially to be part of a newly created stealth team that would match the likes of the top web scraping companies.
As many of you guys know, I like reversing anti-bots, so much that I've barely used browsers to bypass anti-bots as my strength lies in reversing Javascript protections.
When I joined Browserbase I had to roll up my sleeves and dig down into the Chromium source(something I've never done before) to patch all the leaks that Playwright, Selenium and Puppeteer leave.
@pk_iv specifically wanted to meet the customer where they are and not force them to use any custom frameworks. That meant that the only way forward was to build our own custom chrome binaries.
The ugly truth is that there is not a single entity where you can just license or buy all the stealth tools needed to effectively bypass a website's captcha and antibot protection. You have to build a lot of in-house and keep playing the cat-and-mouse game.
When I originally joined my contract was to be remote for the initial 4-6 weeks, but after 4 days I decided to say screw it and drive down to San Francisco and just stay at a hostel near by so I can synchronize better with the team.
I live in Davis,CA which is technically very close to San Francisco, but it is a nightmare to commute every day to be there at 9 AM in the office.
That was the best decision ever as it wouldn't have worked out since a lot of my work needed to be integrated into the Browserbase product.
We use Slack for communication, but waiting for a response back when you can just get help from your co-worker by going to their desk or scheduling a conference room is so much better.
Now regarding @pk_iv , he doesn't have a personal office, desk or stationary place at the office where he works, yet he is there 6-7 days a week.
He eats, lives, and breathes @browserbase and he is everywhere. I'm still amazed at the level of context he is able to hold as he moves from product to engineering to customer support all within a short time frame.
He is answering customer queries, drafting documents on notion, pushing code to github, reviewing PRs, bringing in new customers.
You can't barely see me in the picture, but I'm the guy on the very back with the white board in-front. The guys on the front are all working on stagehand.
Yes, I do have a kid, but he is 7(not 9 @pk_iv lol) going to be 8 next year. I do get to spent time with him on the weekends grinding @fortnite on the Switch.
Yeah, I know I work a lot but honestly I've grown so much in the past 5 weeks that I'm excited to see what the future holds.
The hardest thing for me was working in a team where there are processes you have to follow, discussions that must be had with all the stake holders(people in the team), stand-up meetings where you report what you have done. I've always been sort-of a lone wolf doing my own thing.
@pk_iv is currently looking for more amazing people to add to the team. He hires extremely fast though, you won't wait weeks or months to know if you are a good fit. I was hired under 1 week(but I was referred internally though).
Don't be afraid though, @pk_iv, won't force you or make you work weekends. I actually did it on my own accord and I didn't mind at all as I was able to get way more help(I was stuck on the dev environment) from @pk_iv on the weekends as there were barely any people there.
You will be provided with a:
- MacBook Pro M3
- Unlimited Celsius, cold brew drinks, coconut water and sparking water
- Unlimited snacks(beef jerky, bananas, nuts, goldfish crackers, popcorn)
- Free lunch via DoorDash on Friday
- And last but not least, tons of natural sunlight and beautiful plants that are nurtured and watered by our dedicated PlantLord that comes every week to keep them live and green.
xitter.cf/pk_iv/status/186076206…
Warfa Jibril retweeted
CobolCopilot is building language models that help the world’s largest banks and fintechs make sense of their legacy COBOL systems. Over 30,000 enterprises collectively rely on over 800B lines of COBOL code, and yet it’s nearly impossible to hire a COBOL developer (the average COBOL dev age is almost 60!).
cobolcopilot.notion.site/Fou…
Warfa Jibril retweeted
Honored to moderate a panel discussion with humanitarian operators and aircraft manufacturers at the 16th GHAC today. Our focus was on establishing a collaborative working group between aviation operators and manufacturers.
Congratulations to Paul and the team @browserbase!
Warfa Jibril retweeted
the best way to get good at something is usually to just practice actually doing the thing in question.
a lot of very capable people outsmart themselves with complex plans that involve working a lot on fake prerequisites.
