Harry Tormey 🇮🇪 | 🇺🇸| 🇺🇦
@htormey
CTO 🔨Eliminating toil with AI @stepchangelabs. Full stack, Python, AI, Native iOS & #ReactNative developer. EX: Eng @coinbase @apple @facebook. ☘️ 🏠 🇺🇸
Friday Demo: APX + Datadog catching missing database indexes! Found endpoint with 1-minute responses (39% in DB) & provided exact SQL fixes. Sometimes the biggest wins are hiding in plain sight. Try APX today, link below.
You should be keeping an eye on @cline (open source coder bot) and Kimi. Not that hard to setup with VSCode and @GroqInc
I'd like to point out that for the real world tasks (not benchmarks), Kimi K2 outperforms Gemini. This is telemetry across all @cline users, showing diff edit failure rate. Notice how Kimi has about a 6% failure rate, which is significantly better than Gemini's ~ 10% error…
Google has been shipping some awesome stuff lately but devx still needs some work. This is a venture backed startup in their startups program.
Finally got Gemini-Cli working. Perhaps this helps someone else: 🚫 Work account – Google Auth didn’t work 🔑 Personal account – Google API key didn’t work 🔑 Work account – Google API key didn’t work 🛑 Work GCP + Vertex AI API key didn’t work ✅ Personal Google account –…
I think Tim Apple needs to go shopping and buy a few AI companies. It’s time to start deploying that cash horde.
Verifying AI agents’ outputs is a huge business opportunity. Generic benchmarks like SWEBench have their place but are easy to game and often miss the mark for your use case. We’re focused on how to validate agents as they fix production systems.
AI PROMPTING → AI VERIFYING AI prompting scales, because prompting is just typing. But AI verifying doesn’t scale, because verifying AI output involves much more than just typing. Sometimes you can verify by eye, which is why AI is great for frontend, images, and video. But…
Using @FAL + Claude Code to programmatically create YouTube intros. Burning cash in the best way possible. Video coming soon.
Star Trek TNG’s tech vision has aged well: everyone talks to a superpowered LLM, predatory mobile software is banned, interfaces are minimal and non-distracting, and people work on hard tech problems, not adtech or SaaS dark patterns. Software just exists in the background.
Built a YouTube thumbnail generator using two stacks: Claude Code vs @cline + Kimi (on @GroqInc). Cline stood out, open source, clean dev UX, transparent pricing. Groq + Kimi was fast and cheaper than Claude Sonnet. If Cursor’s getting too pricey, this setup is worth a look.
What’s up with all this Twitter DM group spam? Is there a way to block people adding me to random group chats like this?
This is pretty cool.
GitHub Copilot coding agent just got a major upgrade. ✨ What's new: • It tests its own UI changes with Playwright and adds screenshots to PRs. • It can connect to more context and tools with remote MCP support. • You can trigger and track tasks from a new dashboard. • It…
My co-founder Niall does a great real world walkthrough of what it’s like to use gemini-cli for daily development tasks. @claude_code is clearly the better product. Sadly the whole experience of getting gemini cli setup was super janky (API, authentication, rate limits etc).
So much @Google gemini-cli hype, yet @claude_code CRUSHES it in my real-world tests. Google have a TON of work to do to catch up with Anthropic.
🫡 RIP @mikeal
RIP to a real legend of open-source and Node.js @mikeal As he would, he appropriately over-engineered his memory beyond death and solicited pull requests. Here is mine. While I'm not sure further PRs can be merged, I'm sure they are still appreciated github.com/mikeal/cancer-…
Man, what a contrast between GoogleIO and WWDC this year… All of Apples features thus far seem very meh. Basically a tax on anyone maintaining an existing application versus something that adds value to customers.
Also came here on an F-1 student visa. I feel for anyone going through this nonsense.
I came to the US on an F-1 student visa. Since then, I’ve built a company worth over $20 billion—headquartered in Pennsylvania. Student visas are one of America’s greatest talent funnels. Let’s not close the door behind us.
Want Codex (or any AI model) to help you write better code? Make your modules small and your tests fast. @gdb reminds us: the way we structure codebases changes how well AI can work with them. Amazing interview with @swyx at AIEngineer worlds fair.
Vibes at AIEngineer worlds fair expo. As you can imagine, lots of coder bots.
San Francisco, the intersection of AI, Finance and Crypto.
GitHub coding agents seems pretty locked down: no direct internet connection, you can’t directly kick off a coding agents on behalf of an organization via apps. Interesting seeing how they leverage MCP servers.
