adam_patarino

7 ms·

31.
by adam_patarino 6mo ago
Actually, we're more broadly trained than most models. We did long tail training across languages, so we improved execution with languages like java, swift, and even cobol. It's definitely a david vs goliath. But we know there&#x2
32.
by adam_patarino 6mo ago
It’s not a query / prompt thing though is it? No matter the input LLMs rely on some degree of random. That’s what makes them what they are. We are just trying to force them into deterministic execution which goes against their nature.
33.
by adam_patarino 6mo ago
I’ve lost trust in anything they say. The fear marketing is clearly intentional at this point.
34.
by adam_patarino 6mo ago
Fair enough! The simple answer is: we did a lot of work to make the model better at coding without requiring complicated installation or configuration. One comman to install and run. All the benefits of claude code, without any of the limit
35.
by adam_patarino 7mo ago
Its batteries included. No config. We also fine tuned and did RL on our model, developed a custom context engine, trained an embedding model, and modified MLX to improve inference. Everything is built to work with each other. So it’s more l
36.
by adam_patarino 7mo ago
Tell me more! Thanks for the waitlist
37.
by adam_patarino 7mo ago
I’ve seen this story making the rounds and I’m not just why it’s gotten so much traction. Is it just a good write up?
38.
by adam_patarino 8mo ago
You check the plan files into git? Don’t you end up with dozens of md files? I’ve been copying and pasting the plan into the linear issue or PR to save it, but keep my codebase clean.
39.
by adam_patarino 8mo ago
The biggest gaps are not in hardware or model size. There is a lot of logical fallacy in the industry. Most people believe bigger is better. For model size, compute, tools, etc. The reality in ML is that small models can perform better at a
40.
by adam_patarino 9mo ago
AI has more training on web apps I think the real answer is that these guys are told they need to ship in 2 months and have a huge team of web devs
41.
by adam_patarino 9mo ago
I feel like this is the silent majority. All the twitter hype is not representative of the real world.
42.
by adam_patarino 9mo ago
Yeah but you're still using anthropic's subscription and tokens. That's not really an alternative. That's why we're shipping our own model with cortex.build
43.
by adam_patarino 9mo ago
I am trying so hard to understand wtf people are excited about. I have failed. Claude Code can run over-night or while I'm out. Clawdbot looks like a great way to set tokens on fire.
44.
by adam_patarino 9mo ago
ahem ... cortex.build Current test version runs in 8GB @ 60tks. Lmk if you want to join our early tester group!
45.
by adam_patarino 9mo ago
Yeah, I’ve hit this too. Once you do real agentic work or TDD, you’re optimizing context instead of code. That frustration is why we built Cortex: flat cost, no turn limits, runs locally, and git-aware context so you can just keep going. co
46.
by adam_patarino 9mo ago
That's interesting. I use those moments to show it what not to do. Does it not just repeat the mistakes?
47.
by adam_patarino 9mo ago
Can’t you use git for that? I do that often and just revert changes. It does require me to commit often but that’s probably good anyways.
48.
by adam_patarino 9mo ago
I’ve never understood checkpoints / forks. When do you use them?
49.
by adam_patarino 9mo ago
This is why we HAVE to have a local option and why we're building cortex.build. It's based on a small language model we trained exlusively for coding. We combine it with tools and a context graph designed to be more consistent th
50.
by adam_patarino 9mo ago
Tell me someone has turned this into a library we can use
51.
by adam_patarino 9mo ago
What frustrates me is that Anthropic brags they built cowork in 10 days. They don’t show the seriousness or care required for a product that has access to my data.
52.
by adam_patarino 9mo ago
We build our fine tuning and reinforcement pipeline at cortex.build by synthesizing interactions between a user, the agent loops, and a codebase. The exact data they get from users in Claude Code. That data is critical to improve tool call
53.
by adam_patarino 9mo ago
They are subsidizing Claude code so they can use your data to train better coding models. You’re paying them to show their models how to code better.
54.
by adam_patarino 9mo ago
I think you just wrote a treatment for the next HBO Max sunday drama
55.
by adam_patarino 9mo ago
There are lots of companies that do this. Doesn't make it right. The real "evil" here is that companies like Meta, Google, and now OpenAI sell people a product or service that the customer thinks is the full transaction. I se
56.
by adam_patarino 9mo ago
This. It’s the same play with their browser. They are building the most comprehensive data profile on their users and people are paying them to do it.
57.
by adam_patarino 9mo ago
Proper sandboxing can fix this. But I didn’t see op mention it which I thought was weird
58.
by adam_patarino 9mo ago
I agree with you. I don’t think number of startups or less reliance on funding is the measure though. Businesses are not code. They solve problems, find their customers, convince them to buy their solution, and maintain that relationship. C
59.
by adam_patarino 10mo ago
Feel this so hard. The opposite is also true where you have a micro-service architecture and cursor faceplants in workspaces with multiple repos. We ended up building cortex.build partly because of this exact pain. Our context engine builds
60.
by adam_patarino 10mo ago
Context / understanding of your codebase is the most important part of AI coding. The current AI tools like claude code and cursor aren't super great at this on their own. They want you to use a plan mode to build up a clear plan

More ›