Ravindu Nanayakkara

[Ravindu Nanayakkara @Ravindu_94]

@Southclaws tinfoil theory: The initial spec (particularly elicitation) was purely vibed by 3 series models and Ant just went with it.

[Ravindu Nanayakkara @Ravindu_94]

@mattpocockuk Harnesses should stay thin and generalised IMO. Plus how would this approach scale to other languages?

[Ravindu Nanayakkara @Ravindu_94]

Got Claude to do a little self-reflection.
https://t.co/3pwF6iTM5M

[Ravindu Nanayakkara @Ravindu_94]

@DanielLockyer It wasn't bitter lesson pilled.

[Ravindu Nanayakkara @Ravindu_94]

theo was mid harnessmaxxing when a group of perfoids tweeted and spiked his cortisol levels

[Theo - t3.gg @theo]

“Why did you use Electron for T3 Code?”

Because I evaluated every option, and Electron had the best performance across all platforms for most things.

[Ravindu Nanayakkara @Ravindu_94]

@bcherny Hi Boris!

Sorry to bother on Christmas, but PDF support is quite lackluster when CC/Claude Agent SDK is used with AWS bedrock models instead of Anthropic APIs directly.

IMO, This hinders the dream of using CC as a "general" agent on enterprise data.

https://t.co/MNu9FseTEk

[Ravindu Nanayakkara @Ravindu_94]

@tannerlinsley mfw I spent the last month doing this manually with redis streams and postgres only for it to be released as an HTTP native solution

[Ravindu Nanayakkara @Ravindu_94]

@trq212 Eager to try it. but my existing product heavily uses cwd and agent definitions. The v2 types doesn't seem to support it yet.

[Justin Murphy @jmrphy]

Claude Code is personal AGI. You can't use this thing for more than a weekend without realizing it's completely over. At first you make a GUI app, OK cool. Then you're like wait, GUIs are a waste of time, let's just make a terminal app. Then you're like wait APPS are a drag, what if I just ask Claude Code to do the thing directly? Works immediately. Then you're like damn, now asking Claude Code to do stuff feels like a drag, can I... have Claude make a system that says this stuff for me, in the order I've been saying it, for the reasons I've been saying it? Maybe it can do all my tab changing and clicking too? And you really think you're going to break it by asking for something ridiculous, and then even this works, and that's when you realize.... It's over. Claude is now building you an agent system, and it works. Recursively self-improving machine intelligence, today, on your laptop. An agent system building you a custom agent system which gets better simply by using it (and lightly nudging it to improve based on what was learned in the session). I was bearish on Agents for a while. Not anymore. I'm talking home-cooked retard agents. My wife surely thinks I've lost it once and for all. You have to actually pull yourself away because every attempt to find its limit fails to hit any limit. It's not even 'addictive', it's just not believable that the next crazy idea could also work. And then it works and you're like f*ck... It's over.

[Ravindu Nanayakkara @Ravindu_94]

@thsottiaux Codex Agent SDK.

[Ravindu Nanayakkara @Ravindu_94]

Almost AGI..
#nanobananapro 🍌

[Ravindu Nanayakkara @Ravindu_94]

@davidfowl Is the MCP a wrapper over aspire CLI? Or has extra capabilities?

[Ravindu Nanayakkara @Ravindu_94]

@mitsuhiko My caveman method is asking the agent to clone the libraries I need it to be familiar with to a gitignored subdirectory and generate an https://t.co/NcdhdKviM2 acting as a TOC for followup sessions. Works much better than context7.

[Ravindu Nanayakkara @Ravindu_94]

@wesbos Impressive. Is it more like a voice note that you can reply to or can you have a back and forth conversation interrupting the agent?

[Ravindu Nanayakkara @Ravindu_94]

@simonw Skills should be the next MCP primitive (after tools, resources and prompts) which teaches the LLM to get the best out of the tools at runtime. Much better than the half-baked `serverInstructions`.

[Ravindu Nanayakkara @Ravindu_94]

@davidfowl prompt-first, general purpose agent with full multimodality running at the os level.

[Ravindu Nanayakkara @Ravindu_94]

Has there been any actual research done on whether Spec-Driven Development (SDD) works better over standard iterative prompting with a coding agent?

Some of the so called "specifications" outputted by popular frameworks just feel like poisonous context salad to me.

[Ravindu Nanayakkara @Ravindu_94]

@uhzvs

[Ravindu Nanayakkara @Ravindu_94]

Copilot CLI with gpt-5 is FAST.
Github cooked.

[Ravindu Nanayakkara @Ravindu_94]

@alexalbert__ Yes. The agentic loop is rock solid. MaxTurns is a bit of an issue though.