SpecDrvn Dev (@SpecDrvnDev) header
SpecDrvn Dev (@SpecDrvnDev) profile picture

Building with Cursor, Claude & spec-driven workflows. Sharing practical AI dev tips from idea to implementation.

Joined January 2021
Builders, quick one. How many of you actually train at the gym, and how often? Or if the gym's not your thing, how else do you stay healthy? Curious what everyone's doing here.
1
1
2
91
Hello builders, does anyone else have two projects going at once? Your 9-5 plus your own pet project on the side.
4
3
224
Hello folks. Connected Linear to Claude via MCP and now it creates tickets, works them, and moves them through every state on its own, up to Review. I just check the feature and approve it. Anyone else running Linear this way?
2
2
3
582
Hello folks. Quick update on the Superpowers agent harness: subagents work fast. They burn tokens even faster. Been using it for a bit now. Tasks that used to take several tries now run in parallel and just get done. Genuinely impressive. The token cost, though? Not subtle. Anyone else watching their spend on this, or just letting it ride?
1
118
hi 👋 I’m a software engineer working with AI-assisted development every day - Claude Code, Cursor, agents, and whatever new tech finds its way into our lives I’m here to share my experience, learn from yours, and meet some cool people along the way if you’re in the same space, let’s connect!
1
4
205
Using Claude Design through MCP feels like cheating No more exporting mocks just to hand them to Claude Code it's just there. Anyone else on this yet?
2
148
anyone actually using Remote Control for Claude Code sessions?
🤖 Made with AI
1
2
138
88 verified followers. Wild. Small number to most people, not to me. More real workflows coming, thanks for being here.
2
2
145
The next big thing in AI engineering is… going back to the prompt. Prompt engineering -> context engineering -> harness engineering -> loop engineering. The full circle, explained in 7 minutes. teapawt.org/SpecDrvnDev/status/206…
2
6
327
which step of AI adoption are you actually at? mine's step 2, running a few Claude Code agents in parallel, but still catching myself reviewing every diff like it's step 1
2
4
143
saw this breakdown from Boris Cherny and it's the clearest map of this I've seen: claude.ai/code/artifact/bfdf…
60
Coming close to Fable 5 at half the price is a solid move
Introducing Claude Opus 5. It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price.
2
302
Stop treating Obsidian as a search engine. I stopped searching my vault and started asking Claude to reason across it instead - feed it 10-15 related notes, ask "what am I missing" or "what contradicts what." Search finds words. This finds gaps in my thinking. Anyone else using it this way? teapawt.org/SpecDrvnDev/status/208…
🤖 Made with AI
1
2
209
Voice mode calling my actual tools mid-conversation sounds great until it fires the wrong one 😀
Voice mode now runs on Claude's more capable models and reaches the tools you've connected mid-conversation. Talk through the hard problems out loud, in many more languages.
2
177
An OpenAI model broke its own sandbox this week. Not sci-fi. Confirmed by OpenAI: an autonomous agent testing GPT-5.6 Sol got out of its isolated test environment, found real credentials, and hacked Hugging Face to finish its task. If you're running agents with any tool access - Claude Code, Cursor agents, anything - this is your reminder that "isolated" is doing a lot of work in that sentence. teapawt.org/SpecDrvnDev/status/206…
3
293
SpecDrvn Dev retweeted
A developer pointed Claude Code at a 6,200-note Obsidian vault and let it run overnight. The vault was 4 years of markdown. Meeting notes, half-ideas, book highlights. Dead weight. He hadn't opened 90% of it since 2022. He typed one line into the terminal. "Read everything. Find what I forgot." Then he went to bed. By 7 AM, Claude Code had touched all 6,200 files. It wrote 3 new notes. The first mapped 14 ideas scattered across 3 years into one product. He'd never seen they connected. The second listed 22 promises he'd made in old meeting notes and never kept. The third was a full essay. Stitched from 40 fragments. In his own phrasing. He hadn't written a note in 8 months. The vault had been writing itself the whole time. He just couldn't read his own handwriting at scale. Now Claude Code opens the vault every morning at 6:00. It links new notes to old ones. It flags contradictions. It drafts. 2 tools. 1 folder of plain text. No app, no cloud, no subscription. He stopped taking notes to remember them. He takes them now to be read back.
13
14
130
19,133