How do I view my old posts on @hey? I can see replies going years back but not my feed and I am trying to retrieve an old post I made. How do I do so?

There seem to be a lot of illegitimate SOC 2 certifications happening at the moment. If someone can't tell you who performed the audit, it's not legitimate. In case you missed it, a Y Combinator startup called Delve was completely faking reports (https://deepdelver.substack.com/p/delve-fake-compliance-as-a-service) using Indian certification mills and faking information for the whole process. They were the worst offender, but imo it really boils down to this: is someone disclosing the licensed CPA firm that conducted the audit?If they are not they are lying, it's that simple. They will say who made the report is under NDA, but there is an important difference between the full report, and knowing who conducted the audit. It doesn't make sense to pay a licensed reputable firm and then refuse to disclose who that was. This would also be completely unacceptable in DeFi for very common sense reasons.Of the services you use that claim to be SOC 2 certified (especially as a founder), you should actually check how many disclose who performed the audit. It's kind of shocking how uncommon it is for companies to disclose that obviously necessary information.I'm not an expert on the topic, but I don't think you need to be to see what's happening across the vast majority of small/medium sized orgs claiming SOC 2 compliance

I think one of the things that will differentiate those that are successful with LLMs for coding in the coming years will come from their ability to build up their overall codebase as a continuation of itself. Meaning moving one project forward always means creating a really solid foundation for a future project with minimal duplicated effort that everyone else would likely have to struggle through to arrive to the same place

Sorry for being so quiet on here. I have been really busy building agentic tools and I am now at the point where I will start sharing more interesting things soon. One small update shared here, but I'll start posting more interesting content on here again soon: https://x.com/rickydata42/status/1935066021079265340

Any AI newsletters out there that adapt to what you end up being most interested in and refines itself in the news it brings you?

I'm probably dumb for not knowing this until Claude taught me today, but can run a pre-commit git hook. Useful for catching large files that you might have not caught in your .gitignore, and avoiding a situation where you can't push a commit due to large file size. No dependencies, this is a git feature, you could save the below in `.git/hooks/pre-commit`, which would make sure to exclude any files above 50MB when you make your commits:
```
# Check for files larger than 50MB
large_files=$(git diff --cached --name-only --diff-filter=A | xargs -I {} find {} -size +50M 2>/dev/null)
if [ -n "$large_files" ]; then
echo "Error: The following files are larger than 50MB and cannot be committed:"
echo "$large_files"
echo "Please remove these files or add them to .gitignore"
exit 1
fi
```

https://x.com/CambrianNetwork/status/1906745812316557804

Nothing like getting short glimpes of "agentic flow state" - when an agentic workflow has all the right information feedback loops and ability to keep making progress on a problem iteratively

The new Google Gemini 2.5 Pro model is really impressive, and the next great unlock for agentic coding. Without the paid API having been launched we're back in the era of having access to the best coding model with a chat interface, but can't really take full advantage and not integrated the way sonnet 3.7 is supported everywhere. But in terms of capabilities to me it feels like a jump similar to what sonnet 3.5 had originally given us compared to prior models for coding.It's interesting seeing the large company strategies develop and navigate the open source competitiveness they're facing as well. OpenAI seems to be going more for a platform, and they seem to understand that if they don't embrace open source standards like MCP the way they did on their agents SDK (https://openai.github.io/openai-agents-python/mcp/ ) they won't succeed, despite their competition developing it. I like this approach, and as a result their agents sdk and realtime APIs are things I see a lot of value in using for my work. Anthropic has also been able to position itself well with devs, and carving out some advantages with their Claude Code CLI doing things like very effective sub-task delegation, and solid frequent updates happening there. Google is where I think it gets more interesting. It feels like in the end Google will be able to leverage huge infra with TPUs and everything they have to always have the best models. Generally feels like they have less competitive risk from open source because they can use that to make better models more easily than others, and aren't necessarily having to establish a large user base with devs through API calls, as much as their own path of integrating their existing products/services and devices to their models capabilities.But based on the different directions we see the three giants going in I'm very grateful we have such a highly competitive environment and the pressure that got applied by the chinese open source community in particular and all the cascading effects those catalysts are having and will keep having

It's crazy the length twitter goes to in order to shove Elon content down your throat. I kept picking the option " Not interested in Elon Musk", yet it still consistently shows Elon's posts at the top of the feed every time