tuning agents to build software. 2x apple ssc winner. sih ‘19 finalist.
San Francisco, CA
Joined February 2020
- Tweets1.5K
- Following1.9K
- Followers593
- Likes3.4K
Pinned Tweet
Built wishper, your free alternative Wispr Flow
Download now, Stop Typing and Start Speaking.
free · open-source · on-device · mlx powered
wishper.irangareddy.in/
3
10
1,149
does anyone in my timeline follows weekly releases of software products apps, and any such. what you're learning after claude, codex maxxing
1
2
101
this is what happens when you truly want system act on immediately, good ship @vercel
So bullish on MCP. Its utility exceeds the "M" part: even AI aside, it'll be a civilization-scale accomplishment to get all our systems communicating through one universal open protocol.
A missing piece was how to get arbitrary callbacks across sessions. MCP could read and write, but it couldn't tell you later "this event happened." MCP Events fixes this.
We're adding support in our own MCP. Say an agent makes a deployment, and wants to know when it's complete. Instead of polling, it can subscribe to 𝚍𝚎𝚙𝚕𝚘𝚢𝚖𝚎𝚗𝚝.𝚜𝚞𝚌𝚌𝚎𝚎𝚍𝚎𝚍.
PSA: With Vercel Queues and Workflows, writing the backend machinery for your own MCPs to do this is easy peasy.
44
got 3 👀
We're giving Ultra and Team Premium users Fable resets based on how long you've subscribed:
<6 mo: 1
6–12 mo: 2
>12 mo: 3
Use them anytime by March 8.
Pro and Team Standard get a week to try Fable 5.2, up to half your weekly limit, plus one reset. support.claude.com/en/articl…
2
50
/goal automate workflow
agents orchestrations:
nitter.cf/ryaneshea/status/21187…
agent harness:
nitter.cf/joelniklaus/status/211…
I burned a shitload of compute so you don't have to. Releasing the Harness Leaderboard.
Which harness should you run your open model in? I ran 28,849 trials to find out.
The leaderboard compares nine coding-agent harnesses on open models. In each comparison only the harness changes: the model, provider and reasoning effort stay fixed, and every task gets three attempts.
The trials ran GLM-5.3-Flash, DeepSeek-V4.1-Flash and Flint on Terminal-Bench 2.1 and DeepSWE, for $4,746 of inference.
Averaged over the three models, mini-swe-agent (@KLieret), a bash-only agent, solves 51.4% of tasks and @openclaw 27.3%. Four others are statistically tied with mini-swe-agent.
One of them is Pi (@pidotdev), at 49.4% for $0.29 per solved task, about 45% less than mini-swe-agent's $0.53.
The best harness depends on the model: mini-swe-agent on GLM-5.3-Flash, Hermes Agent (@NousResearch) on DeepSeek-V4.1-Flash and Kimi Code (@Kimi_Moonshot) on Flint.
@opencode costs $3.35 per solved task, 11.7 times as much as Pi.
See Harness Hub's Harness Lens for what the nine harnesses do differently. It documents the prompts, tools and compaction of 34 coding-agent harnesses and links every claim to its source.
Leaderboard: huggingface.co/spaces/joelni…
Harness Lens: huggingface.co/spaces/harnes…
42
ci/cd costs are growing faster as agentic coding increases commit and pr frequency, triggering more builds, tests, and deployments.
going back to first principles: identify redundant runs, optimize caching, run only affected tests, and reduce unnecessary compute without compromising reliability.
building a tool for small teams to analyze ci/cd workflows and cut costs.
my weekend is locked.
will mostly find a better skill for all repos we have backend api, iOS, Android, and Web(React) or create one.
5
2
155
one thing that caught my eye this week is
nitter.cf/ShopifyEng/status/2118…
We rebuilt our mobile apps in Swift and Kotlin with agents. To keep pace and hold our quality bar, we built Reflex: an agentic-first architecture that lets agents build and verify app behavior across iOS and Android without needing devices or simulators.
Here's how it works, piece by piece:
shopify.engineering/reflex
18
Ranga Reddy retweeted
Apple's Core AI models repo is now accepting third-party contributions 🥳
If you have developed some cool ports, this is the best place to share your recipes with the community. Looking at you @JackdeS11 @mattt @kevinqz 😂
github.com/apple/coreai-mode…
Ranga Reddy retweeted
18 of the 20 best open models today are Mixture-of-Experts. Dense is basically over.
And yet the know-how to train a good MoE still lives mostly in the heads of a handful of teams.
So we spent 800k GPU hours running the experiments ourselves and wrote everything down. Super happy to finally release "The Ultimate MoE Training Handbook", free and open, organized around the questions training teams hit:
- how many experts, how big, shared or not?
- why is half my step spent in all-to-all, and how do I hide it?
- why is my run 25x slower with no error in the logs?
- softmax, sigmoid or sqrt(softplus) router, and why keep it in fp32?
- aux loss or DeepSeek's loss-free bias?
- upcycle a dense checkpoint or train from scratch?
- can I serve with fewer experts per token than I trained with?
⚠️ Warning: it's long (~70k words, budget a few days, sorry) but with 55 interactive figures, code you can run and numbers from our own runs.
Read it here: huggingface.co/spaces/huggin…
Lately me :)
Andrej Karpathy’s setup from his HackerTube blog today:
- “I use Claude Code (desktop app for Mac).”
As I’ve been saying, terminal is the wrong interface for coding agents. Desktop apps are the best interface today.
- “I use Opus 5.5 on Medium effort as my default.”
There’s no point to use the smartest model at the highest reasoning effort for most tasks.
50
i'm a codex power user, gave grok's $60 plan a shot to sweep some platform ui work in pstack mode, and it actually worked pretty well. wasn't a big fan of persistent chats/agents until claude with opus 5.5 changed my mind. they actually stay on track, act well and keep things clear. now i'm mostly using claude and codex, so figured i'd downgrade grok to the $30 plan. there was still an option to continue at $60 for six months, but literally two days later it's gone and everything's bumped to $100 lol.
if you're like me, using claude to steer and codex to do the work, or have found any better ways to steer agents, let me know. definitely down to give them a try.
1
83
I went too far with a design prompt for figma, and opus basically said, if company x and company y had a baby with these product features, this would be it.
184
If opus 5.5’s efficiency and quality hold with no downgrades, single-digit teams can build, ship and scale serious tech businesses, not just SaaS but full consumer platforms.
46
codex fixed fast, issue was an iOS Simulator clipboard-sync loop.
running simulators were repeatedly overwriting the Mac’s clipboard several times per second. So ⌘C and ⌘V were detected, but the copied content became unavailable before it could paste.
It disabled automatic Simulator clipboard syncing and stopped its active sync helpers. The clipboard stabilized, and copy/paste worked in TextEdit.
1
1
66
Companies using agents at scale have same problem, everyone needs there own git
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
46
