Kumara Krishnan

The last 20% isn't most of the work, it's all of the work.

It's all fun and games until you need to implement auth...

If Bengaluru is Silicon Valley, Hosekote is Texas. New-age industrial founders, all locked in, long Indian dynamism 🖖🏻

Github Copilot in Visual Studio 2026 is such a mess!! It just can't run edit operation on larger file. It always miss the edit context and apply the edit in the wrong place.

Tip: Figure out your personal ceiling for running multiple agents in parallel. We need to accept that more agents running doesn't mean more of _you_ available. The narrative is still mostly about throughput and parallelism, but almost nobody's talking about what it actually costs the human in the loop. You're holding multiple problem contexts in your head at once, making judgment calls continuously, and absorbing the anxiety of not knowing what any one agent might be quietly getting wrong. That's a new kind of cognitive labor we don't have good language for yet. I've started treating long agentic sessions the way I'd treat deep focus work: time-boxed and tighter scopes per agent dramatically change how much mental overhead each thread carries. Finding your personal ceiling with these tools is itself a skill and most of us are going to learn it the hard way before we learn it intentionally.

Ever visit a GitHub Issue and wonder when the fix shipped? Release info is now directly included in the issue sidebar alongside the linked PR.

It used to be universal truth that it was absolutely not worth spending time and effort building software that was not your core competency With building being cheap, rebuilding internal tools is actually easier and also fun + builds the AI muscle But maintaining still remains a pain that is devs have historically been wonderful at underestimating.

Cannot wait for teams that build their custom JIRA/Workday replacement/custom CRM to one day turn around and ask: “Why do we have so much internal software that is buggy / has poor UX + we need to maitain?” Seen this movie well before AI, when Uber built uChat (custom Slack)

“not knowing how to cook is becoming an advantage” - ceo of mcdonalds

Too many people are becoming too helplessly dependent on these tools before it’s even clear what the real, eventual cost to consumers will be.

@sanjaymaniam Both, xml comments in code into api documentation. md articles grouped together into a separate section.

@_KumaraKrishnan Nice! Did you use it to create documentation from source code / web page from markdown?

DocFX is seriously impressive. Building an MCP docs server on top of it was a breeze.

defuddle github.com/kepano/defuddle

While everyone's talking about Claude Code Max $200/month subscriptions, there's GitHub Copilot Pro+ which only costs $39/month with *1500 premium requests included*. I swear what I'm doing with it costs a *LOT* more than $39 after burning $120 in 24 hours with Cursor

@championswimmer Tbh, both the Pro and Pro+ plans are a steal. Esp the Pro+ which offers 1500 premium requests, along with 375 spark credits as well. One could use the free models to refine their prompts to ready a spec in plan mode and have premium models implement it in agent.

@jsensarma There’s some model releasing every month so something is always free. Take what you get when you get I guess Also GitHub Copilot $100/yr plan is incredible value and has all models available on it. And there is a $39/month plan too which I can guarantee 99% people can’t exhaust

one thing not being discussed about AI in India - we need more cheap AI credits for devs. $200/month is very small sum for American dev. even $1k/month. but for most Indian firms/devs, these are larg-ish sums. leave alone students/hobbyists. this will set us back skill-wise.

Message steering and queueing just landed in today's @code stable release. Other notable features in the release: agent hooks, Claude compatibility, and skills as slash commands. See them demoed live on Let It Cook tomorrow: youtube.com/live/FjvtWeG6E…

Imagine waking up to calm... ✅ Issues triaged ✅ CI failures investigated + fixes ✅ 2 new PRs improving your tests Chores done, problems solved 🪄 Join us in shaping the future of repository automation with GitHub Agentic Workflows. github.blog/ai-and-ml/auto…

feels like such a wasted opportunity every moment your agents aren't running

💯

rough thoughts on dotnet (csharp) and agentic - agentmd should steer towards latest c# style ie “only use .net 10 code style” because of the wide degree of variance over the years. even with that guardrail in place it’s a mixed bag because the language has evolved so much over the years and models have that in their training data. - migrating a codebase from v8 coding style to v10 coding style super easy to do! - dotnet null just sucks; discriminated unions - miss eithert/maybe heaps and plenty of scenarios which should have been compiler time validation end up as runtime blowups. - writing roslyn analysers is now neat, fun and super productive. there’s no excuse to not utilise them to create back pressure - nuget still sucks; can’t ripgrep a dll and monkey patching libraries still sucks/not really possible/easy to do. - configuring compiler warnings to max level of aggression helps heaps. warnings should fail builds.