Siddhant Sharma

@tonysimons_ Nice article!! soul md files are a core feature in paperclip too. that’s how they define the role of an agent effectively

@RakibullHassa13 that looks insane! well done

it's happening🤩!

Wait, a coding agent that's actually FREE? Not "free trial." Not "free tier with 10 requests." Free as in — install it, use it, never pay. Here's how: ads show up in your terminal. That's the whole trick. The ads fund the models, not you. And the models aren't cheap ones either: → DeepSeek / Kimi K2 / MiniMax handling your code → Gemini Flash for file search → GPT-5 deep thinking if you have ChatGPT Oh and it beats Claude Code. 61% vs 53% on 175+ real coding tasks. We've been paying $20/month for this. npm i -g freebuff (available in select countries — check freebuff.com)

@hasantoxr Great share. This can level up the social media game in so many way. Excited to try out

Holy shit... someone made it work that I can integrate a social media manager inside Claude It's called OmniSocials and it lets Claude draft, schedule, and publish to 10 platforms from a single command. No Hootsuite. No Buffer. No $200/seat pricing. Here's everything you need to know: 👇

@ds3638 Is there a link to the talk?

Evals are dead. Or more precisely: traditional eval-driven development doesn’t scale. Static evals were useful when agents were short-lived and bounded, but once agents are running for hours and taking thousands of actions + operating autonomously, evals alone stop being enough. At that point pass/fail is too coarse. Simulation misses too much of what happens in prod and model capabilities are moving faster than eval infra can keep up. What we run instead: observability-driven development. - deploy with tight guardrails - collect prod trajectories - cluster behavior to discover patterns + failure modes - specialize workers for narrower tasks - tune thresholds until behavior is reliably within bounds Can you see what your agents are doing? Can you detect drifts before they cause damage? This is an important shift in how we build AI systems. Evals still matter but observability is becoming the foundation for prod-ready agents. Thanks Sunny Bakhda (@honeyhiveai founding engineer) for a great talk at @aicouncilconf

@icanvardar here’s a reason x.com/subtlemusing/s…

cancel your claude subscription start using codex thank me later

if you're using claude -p in production right now — reply with what you're using it for and roughly how much of your weekly limit it eats.

the businesses most exposed are exactly the ones who found the most value in -p. small teams. lean pipelines. smart automation built on a tool that felt free because it was already in the plan. that calculus just changed.

Anthropic just announced dedicated monthly credits for programmatic claude usage starting june 15. everyone's calling it a win. but nobody's asking the right question. 🧵

@BathaniDev Congrats on the launch. Exactly what the users need. Do you have an estimate on an average token churn to turn a plan into design?

I'm thrilled to announce that we're launching polygram.dev 🚀 Most AI tools start generating code before understanding what you actually want to build. So we built a different workflow: 🧠 Plan → 🎨 Design → ⚡ Build We're Just getting started👇

@ClaudeDevs what are some use cases where claude -p was actually providing value over the interactive mode. It’s either Claude Code or SDK, no?

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

Prompt caching is the easiest cost win in production AI. I wrote a full breakdown — the math, the silent killers, the provider comparison. x.com/subtlemusing/s… I'm documenting everything I learn about deploying AI at scale. If that's useful, follow along.

Silent killer #2: Minimum Token Threshold Minimums: • Claude Sonnet / Opus: 1,024 tokens • Claude Haiku: 2,048 tokens If your system prompt is 800 tokens — you're paying the cache write fee and getting zero hits.

You added prompt caching to your AI app but still paying full price. No error. No warning. Just silent cache misses. Here's exactly why — and how to fix it. 🧵