Niket Patil

I got tired of flaky tests destroying our Jenkins builds. So one weekend I sat with Claude and built a GitHub Action. It runs your tests 3x, catches the flaky ones, and uses AI to tell you WHY they're flaky. Free. Open source. Just shipped v1. 🔗 github.com/iamniket/flaky…

@GithubProjects github.com/iamniket

|￣￣￣￣￣￣￣￣￣￣￣￣￣￣| | Share your GitHub profile | |＿＿＿＿＿＿＿＿＿＿＿＿＿＿| \ (•◡•) / \ / —— | | |_ |_

@solana The future: AI agents thinking at Claude speed. Decisions executing at Solana speed. Most are still arguing about which is overhyped.

Best replies get a follow

Want to start your career as an AI engineer? I spent weeks putting together a list of exactly what to learn, in what order — based on what's actually being asked at interviews in 2026. DM me "AI" and I'll send it across. (Free. No catch.)

@N_and_ni Google has all the ingredients. OpenAI and Anthropic have the urgency. Big difference.

Google has the data. Google has the talent. Google has the infrastructure So why aren’t they dominating AI? 🤔

5 Claude prompts I use daily at work: "Rewrite so it's not passive aggressive." "Summarise. What did we decide?" "Make this update sound productive." "Strongest counter-argument to this?" "Explain so a tired colleague gets it." Save this. Thank me Monday.

In other news OpenAI just launched a $4 billion consulting subsidiary called DeployCo. Translation: they realised enterprises won't deploy AI on their own. The real AI war isn't "best benchmarks." It's "who controls how Fortune 500s actually USE this stuff."

KPMG just deployed Claude to 276,000 employees across 138 countries. Meanwhile your company is still in a 6-month "AI evaluation committee" meeting. Big 4 firms know what's coming. Most companies will figure it out in 2027.

@rxhit05 Your tokens disappear 20× faster than free snacks in an office pantry.

has anyone tried the new Opus 4.8? how is it?

Genuine question: Why does every AI provider have a completely different way of pricing tokens? OpenAI: $X per million input, $Y per million output Anthropic: same structure, different math Google: cached vs uncached vs grounded xAI: ??? Just give me a flat monthly bill

@_MaxBlade And with single prompt my token limit is gone 😢

your one job today is to open opus 4.8 set /effort to ultracode ask it to make your app 100x better in everyway

Claude Opus 4.8 dropped. Same price as 4.7. → SWE-Bench Pro: 69.2% (beats GPT-5.5 by 10 pts) → OSWorld: 83.4% (best in field) → Humanity's Last Exam: 57.9% (highest) → Terminal-Bench: 74.6% (GPT-5.5 wins at 78.2%) Anthropic ships faster than most teams ship meeting notes.

@Rankdawnn Opus 4.9

Claude opus 4.8 dropped !!! What u are building with it Share below 👇

@mark_k I've never seen a company roll back their model right after release. That's the AI equivalent of a Friday deployment followed by, "Guys, let's pretend this never happened."

Codex release was moved to Friday (today), OpenAI just announced.

😒

@madsmadsdk I recently went for an interview and they asked me to write on paper. For a moment, I thought I had accidentally time-traveled back to 2005.

I wrote code by hand today. Ask me anything.

@Hiteshdotcom I was waiting for this tweet. I kept my entire week quota for this weekend let's see how it goes.

Opus 4.8 is getting limit exceed with just 1 prompt. I guess that’s a sign to take day off😂

Yday I got an email from my manager about overusing Claude tokens. Just checking - while working against urgent deadlines and trying to get things delivered on time, are we now expected to track token consumption as a KPI as well? Should I start adding a "tokens saved" section

@ClaudeDevs Going to try this over the weekend let's see is it worth the hype

Opus 4.8 is especially strong at long-horizon work. In Claude Code, pair it with /goal. If you're building on Claude Managed Agents, try it with Outcomes.

Opus 4.8 is live in Claude Code today. A few things worth knowing: 🧵

Anthropic just dropped Claude Opus 4.8. Highlights: → Codebase-scale migrations end-to-end → Dynamic workflows with subagents → Fast mode 2.5x faster, 3x cheaper Genuine question: how long until your company actually approves it for work? 😅 #claude #opus

Will be testing 4.8 this weekend on my automation stack. Specifically curious about: → How well it holds context across a full Java/Maven repo → If "dynamic workflows" can replace my multi-step CI scripts → Whether the new fast mode actually maintains quality Will share res

Claude Opus 4.8 just dropped. 41 days after 4.7. Same price. Highlights from Anthropic: → Outperforms GPT-5.5 and Gemini 3.1 Pro on agentic benchmarks → Codebase-scale migrations from kickoff to merge → Fast mode now 2.5x faster AND 3x cheaper → "Dynamic workflows" running