Itβs been an amazing 3 years at @cyfrin.
No, this is not a goodbye post.
Iβve just been reflecting a bit after our latest launch. Over the last 3 years, Iβve had the chance to work on Updraft, Solodit, CodeHawks, and now Cygent, which makes this one especially exciting.
Cygent is a very cool product to be part of. Itβs an AI security engineer built to help teams find vulnerabilities, write fixes, and open PRs, all while working alongside developers instead of just dumping another report on them.
More than anything, Iβm proud of the team behind it. I get to work with a top-notch engineering group that cares deeply about quality, speed, and building useful things the right way.
Very proud of what weβve built so far, and very excited for year 4 at Cyfrin.
If you're not solving harder problems then you were solving a year ago then you're not using AI right.
This is the video.
If you're creating crazy AI workflows and trying to figure out the best prompts to get some crazy output.
Watch this.
Prompting isn't some big brain engineering task. It's a language and steering problem.
Looking for my next full-stack role -> lucas.website
Experience working with multiple Web3 companies, Apple, Amazon, and doing agency work.
I've shipped and scaled production apps from first prototypes to thousands of users across various stacks.
Have experience using the latest AI tools and setting up AI workflows and pipelines.
Bonus: I can still manually read/write code
Craziest shit I'm seeing is that most AI doomers are rewriting history as if most enterprise software pre AI was some incredibly engineered, maintainable, secure and fast codebase.
Brother...AI slop has nothing on a 10 year old strung together piece of shit that was maintained by 7 different outsourced engineering firms.
The biggest threat to the AI boom might be a boring boardroom question:
βWe're spending more on AI. Are we actually making more money?β
If companies can't answer that, what happens when those enterprise contracts come up for renewal?
Mike @ HTML All The Things π¨π¦ retweeted
I feel like open source is moving towards 'shared compute'
If someone wants to offer their tokens to fix well documented issues/add needed features then maintainers can open specific issues for that type of contribution.
Obviously still needs someone that is good at steering but I can see the direction heading this way.
I feel like open source is moving towards 'shared compute'
If someone wants to offer their tokens to fix well documented issues/add needed features then maintainers can open specific issues for that type of contribution.
Obviously still needs someone that is good at steering but I can see the direction heading this way.
Sticking to as close to base harness as possible just for this exact reason.
As soon as you get comfortable in a custom workflow a new model comes out that destroys it.
Run this before trusting any repo you get sent.
We've gotta be 10x more diligent now that malware creation has become easier then ever.
I've pushed significant updates to coldclone, a very useful tool for sanitizing foreign repos before engaging with them.
If you're an auditor doing a lot of quoting or a dev downloading new repos to use, running this first is a no-brainer github.com/devdacian/coldcloβ¦
I'm more excited for AI to get cheaper and faster then smarter at this point.
I think the companies are starting to prioritize these as well which should hopefully make this whole bubble a bit more sustainable.
So Copilot copilots Copilot and uses Copilot inside the apps and then everything is reviewed by Copilot.
Yup great branding.
Weβre building Copilot as a new OS for work that spans every model, every form factor, and every task. Today, weβre announcing our biggest update to Copilot to date, bringing four things together:
Β· Autopilot: proactive and long-running agent built for the enterprise
Β· Code: build apps with Copilot, hosted inside your companyβs tenant
Β· Home: Chat + Cowork together
Β· Office: now fully embedded in Copilot (and Copilot embedded in Office, of course!)
Plus, you can invoke Copilot in Teams, and weβre introducing Today, a proactive experience that surfaces the most important information from across M365 without needing to ask for it.
The way we work is changing and so are our workflows. This update brings AI into that flow, from answering a question, to building an app, to getting work done on your behalf.
Mike @ HTML All The Things π¨π¦ retweeted
Does it even matter that AI introduces more technical debt into a project if it itself can address and fix it whenever it actually causes an issue?
The biggest unlock I've had recently is giving Codex/Claude safe access to production servers to monitor and repair issues on the fly.
I've been doing it on some small projects just to test it out but holy shit, I really don't think developers will need to be on call anymore in the near future.
These repair and deployment loops have been eye opening.
I genuinely feel bad for people who tried ChatGPT 3.5, got some nonsense back, and wrote off AI entirely.
Does it even matter that AI introduces more technical debt into a project if it itself can address and fix it whenever it actually causes an issue?
I just spent 3 hours with fable creating the perfect feature
sol did a thorough review
grok did a simplicity pass, catching extra guardrails that should not exist and code that is just superfluous
sol did one more pass ensure simplicity didn't break any contracts
and all of this to produce one of the worst interfaces I have ever seen. Actual pure insanity









