Dev Jain

@buildingwwdavid @zaimiri yeah sure i'd be down to see it

@corsairdotdev @zaimiri facts, the error handling alone in n8n can eat an entire afternoon. we're working on something way simpler — runs locally, flat fee, bring your own AI keys. interested in trying it?

Built a bot that watches Client Workflows for n8n. I run workflows for multiple clients in n8n. Sometimes something breaks & I need to fix it before the client finds out Now I just get notified on Telegram and click "Approve" > Spawns headless Claude Code session > Fixes the n8n workflow > All from my phone > Client pays me retainer to keep it working What a life.

@Rudra1071219 Hey @Rudra1071219 ! Sorry I just saw this message! Glad you joined the discord 😄

@corsairdotdev Hey dev, I came across Corsair on GitHub, really like the idea of abstracting away the integration plumbing (that 95% pain is too real 😅). Is it open for external contributions? I’d love to explore contributing, especially around integrations or developer experience.

Tool calling is broken right now, and Corsair fixes it. Corsair lets your agent connect to anything online via APIs. It stores all data in your db so you can access it anytime. Built in webhooks, secure auth handling, strong types, and impossible to use wrong. Add it to your project with `npm i corsair`

@kentcdodds another benefit is you can quickly deactivate the feature if something isn't going to plan. instead of having to wait to rollback changes and redeploy, you can just toggle it off.

I was literally thinking yesterday that feature flags become so much more important with AI implementations. You can more confidently ship features and get them into the hands of early adopter users who are excited to test out things early (even if things aren't nailed down) making it much easier to experiment with new ideas. Cloudflare out here solving real problems!

@zaimiri makes sense! if you want to make it even better, id suggest you try corsair. it’s easier for an agent to work with and builds more complex workflows reliably

@corsairdotdev so life is easier with this new bot now

@ConnerBean silver lining 😂

@corsairdotdev ahhh so this is why i'm so bad at texting nice

the best engineers I've worked with had some of the worst resumes

@enesakar i aim for the most efficient coding agents so i can get nothing done faster

Something’s off when you spend more on time tools than on the work itself.

@aakashgupta agent wars will always be on the API layer. if any product has sufficient demand for an API layer, it has to eventually offer one or risk losing their customer base. cua and browser use are great on the UI layer, but i think they are a bandage solution until that product has api

The framing here tells you everything. "Workflows that don't expose an API." OpenAI spent three years convincing the software industry to build structured integrations. Function calling in 2023. The plugin store. Actions. Then MCP adoption pushed across every major SaaS vendor through 2025. The entire premise was that models work best when apps expose clean interfaces. Shipping a computer-use agent is the admission that premise didn't hold. Most internal enterprise software will never get a modern API. Hospitals still run on systems built in 2004. Insurance claims flow through Windows apps with no URL scheme. Oracle, SAP, Workday, the long tail of government portals. Billions of dollars of labor sit trapped behind software that was never going to expose structured endpoints to an agent. You can pay a sales team to convince Epic to build an MCP server, or you can ship a vision model that clicks buttons. OpenAI just picked the second one. The timing is the other tell. Anthropic shipped computer use in October 2024. They shipped on-device computer use in Cowork and Claude Code on March 24 of this year. That's 23 days ago. OpenAI's response window on a headline capability has compressed from 6 months to 3 weeks. Worth watching what this does to the valuations of every AI integration startup that raised on "we build the connectors." If the model can just drive the mouse, the connector layer gets thinner fast. The bet now is whether the vision-first approach hits the 95% reliability bar enterprise workflows actually need. Claude is around 61% on OSWorld-Verified. Human baseline is 72%. The gap is closing but it's still the bottleneck on the whole thesis. The agent wars just moved from the API layer to the pixel layer.

@zuess05 plumber

Serious question. For the last 10 years, society told everyone "just learn to code" to escape the middle class. Now Claude writes the code. What exactly is the career advice for an 18-year-old right now?

@kylegawley when you don’t see results in the first day or two you assume no one is interested and move on. but tweaking the message is just as important as tweaking the product

problem with shipping fast is you have no opportunity to measure what’s actually moving the needle Experiments take weeks to run

@paularambles market cap of figma is still 50x higher

welcome to the future

@DavidKPiano is this for introspection? kind of letting the agents see what they have access to? not fully following

Released the XState v5 skill 🚀 The goal isn't just "know the API" but rather "help agents make really good state machines" Tuned & tested against real examples, common anti-patterns, and the mistakes* LLMs tend to make with XState npx skills add statelyai/skills

@KaraBharat template literals are a goldmine of typescript tricks

⚡️ TypeScript utility to map nested object keys to a flat union of dot-notation paths #TypeScript

@SDinesh91 and "Observable"

Devs also discovered the word "Durable".

@SDinesh91 when you don't have a result, you make your input the metric lines of code and tokens are inputs

Lines of code were already a bad metric for developer productivity. And now we have an even worse metric popping up "tokens burned". What's wrong with the software industry 🤦‍♂️

@mxkaske ah gotcha, that makes sense! btw, i've used this a few times in my own projects haha great work! a good data table is so underappreciated

@corsairdotdev yep, but also the project exists now for almost 2 years

logs.run just hit 2k ⭐ - appreciate every single one!

@mxkaske Like SEO with Google? I’m trying to rank more on that as well with Corsair but have been having trouble

@corsairdotdev mainly google and twitter

@JussCubs if you can have better MCP context management, you can waste a lot less tokens with little effort

if u are maxing out codex or claude code credits, you are not leveraging MCP server tools, memory, and docs/context management if you feel like you use too many tokens, that's almost certainly your issue your LLM has to pull all new context every run study context management

@gakonst @mpp aren't you seeing a lot of context bloat from adding so many mcp servers?

how to monetize your MCP server? we just published a guide on this using @mpp w/ pretty epic developer experience mpp.dev/guides/monetiz…