🚀💥AI Innovator in #CustomerSupport 🤖 | Driving 80% ticket resolution via #AI | Sharing insights on #AIForSupport & tech transformations.
Joined January 2023
- Tweets 1,407
- Following 269
- Followers 153
- Likes 1,953
59 Photos and videos
Pinned Tweet
24 May 2024
Next-level RAG🤯:
- multiple agents
- web searches
- LLM per-user memory
- basic RAG
I envision this integrated in AI support, with the LLM being able to provide not only responses backed up by real data sources, but also fine-tuned to each customer.
Awesome demo @ashpreetbedi !
22 May 2024
Introducing `Personalized Agentic RAG` - where the LLM remembers key details about the user and automatically chooses the tool for RAG.
We'll build:
đź‘Ź ChatGPT like memory
⚙️ Function calling
🧙 Multi-Agent orchestration
code: git.new/personalization
1
9
621
Ciprian Nastase retweeted
Unpopular opinion: most companies don't need AI agents.
They need to fix their data first.
I've been deep in the trenches building a renewals agent for the last few months. Every single hard problem we hit had nothing to do with AI.
You can throw the smartest model in the world at this. It will still fail. Because the model is reading garbage.
The teams winning with AI right now aren't the ones with the best prompts. They're the ones who quietly cleaned up their CRM, fixed their pipelines, and built proper guardrails BEFORE plugging an agent in.
The boring work is the moat.
If you're a leader thinking about "AI transformation," ask yourself: would a new human hire be able to do this job with the data we have today?
If the answer is no, an AI agent won't save you either.
Fix the plumbing first. Then bring the agent.
2
2
55
Ciprian Nastase retweeted
Apr 20
Gen Z, Gen Alpha, now Gen AI...
One of our AI Innovation leaders messaged me from New Delhi this weekend. His 19-year-old daughter was signing with the Airports Authority of India. Her AI feedback platform is deploying across 100 airports.
She built it in weeks.
We didn't turn her dad into an AI guy. He already was one. That's who we recruit.
What we do is drop world-class talent into a culture that refuses to sit still. Weekly shipping. Demos instead of KPI reviews. A constant sense of urgency that we aren't ahead, we're just not getting run over from behind yet.
That AI DNA goes home with our people. Shows up at the dinner table. Shapes how their kids think about what's buildable.
A 19-year-old shipping production systems at enterprise scale doesn't come out of nowhere. She came from the water this family swims in.
Cool things are truly happening here.
1
9
18
334
Anyone else moving from Cursor (Opus 4.6 driving) to Claude Code and suddenly it feels like 2024 again? I ported my agent instructions expecting minor tweaks, but I’m getting nonstop spaghetti code 🍝
Anyone got tips to smooth the transition?
2
47
Ciprian Nastase retweeted
Mar 24
The old way: sit down and start from zero.
The new way: describe the work once, it is already in motion.
AI is not a shortcut. It is a role shift.
We are not the operators anymore. We are the managers.
Most people have not adjusted.
The ones who have are not working harder.
They just never start from zero.
#AI #GenAI #WorkSmarter #Productivity
1
1
33
388
2,283
16,398
6,898,814
Ciprian Nastase retweeted
Mar 4
There's a built-in agent inside Claude Code you can invoke just by telling Claude to use the claude-code-guide agent
Save this... you'll need it
It's not a local file in your project. It's a built-in subagent defined in Claude Code's system itself.
It answers questions about Claude Code CLI, the Agent SDK, and the Claude API.
It uses Glob, Grep, Read, WebFetch, and WebSearch, but can't edit or write files. Its only job is to answer questions.
26
46
556
57,268
Ciprian Nastase retweeted
Mar 17
Most people are using AI like a toy.
A few ones are using it like a teammate.
Ask it to brainstorm, clean up writing, organize research, and help you think through decisions. Not to do your job for you, but to help you do your job better.
AI works best when you bring the context and judgment.
#AI #GenAI #WorkSmarter #AIForWork
1
1
26
Ciprian Nastase retweeted
Feb 24
Human interaction is going to shift to discords and group chats, invite-only. The open web and social media are going to be left for the agents lurking amongst the ruins. Everything public will be Moltbook.
146
111
1,374
109,946
Ciprian Nastase retweeted
Feb 23
New research: The AI Fluency Index.
We tracked 11 behaviors across thousands of Claude.ai conversations—for example, how often people iterate and refine their work with Claude—to measure how well people collaborate with AI.
Read more: anthropic.com/research/AI-fl…
210
301
2,654
531,024
Ciprian Nastase retweeted
Feb 20
This is completely wrong.
Try the same difficult prompt on an LLM from a year ago and a frontier LLM from today. It's not even close.
Burying your head in the sand is comfortable, but not helpful long-term.
Make up your mind after trying the best AI systems available today.
Feb 20
it's WILD that no one has noticed that LLM progress has stalled. All the signs are there
95
20
452
80,277
Ciprian Nastase retweeted
Feb 17
People should read the Claude Constitution. It does a pretty good job of laying out what Anthropic presumably really believes (and it is part of training). I’d think that a clear debate over things that are good or bad or missing there would be helpful. anthropic.com/constitution
48
23
307
22,868
Ciprian Nastase retweeted
Feb 6
This is literally everything you need to master Claude Code:
93
1,675
11,013
880,689
Ciprian Nastase retweeted
Feb 5
Increasingly believe that the next model after centaurs/cyborgs looks like management of an organization. Decisions flowing up from multiple projects, most handled semi-autonomously, but with strategy, direction, feedback, approval made by the human. Not the final state, though.
it’s just so clear humans are the bottleneck to writing software. number of agents we can manage, information flow, state management. there will just be no centaurs soon as it is not a stable state
19
26
348
59,417
Ciprian Nastase retweeted
Feb 4
Built "Inner Loop Ralph" - a Claude Code plugin where Claude writes the PRD for you.
Describe what you want
→ Claude breaks it into tasks
→ shows you the plan
→ executes autonomously
Works for code, research, writing, etc.
3 commands to install.
github.com/dschwartzi/inner-…
1
2
4
220
Let's see if Sonnet 5 actually releases tomorrow...
Feb 2
Claude Sonnet 5: The “Fennec” Leaks
- Fennec Codename: Leaked internal codename for Claude Sonnet 5, reportedly one full generation ahead of Gemini’s “Snow Bunny.”
- Imminent Release: A Vertex AI error log lists claude-sonnet-5@20260203, pointing to a February 3, 2026 release window.
- Aggressive Pricing: Rumored to be 50% cheaper than Claude Opus 4.5 while outperforming it across metrics.
- Massive Context: Retains the 1M token context window, but runs significantly faster.
- TPU Acceleration: Allegedly trained/optimized on Google TPUs, enabling higher throughput and lower latency.
- Claude Code Evolution: Can spawn specialized sub-agents (backend, QA, researcher) that work in parallel from the terminal.
- “Dev Team” Mode: Agents run autonomously in the background you give a brief, they build the full feature like human teammates.
- Benchmarking Beast: Insider leaks claim it surpasses 80.9% on SWE-Bench, effectively outscoring current coding models.
- Vertex Confirmation: The 404 on the specific Sonnet 5 ID suggests the model already exists in Google’s infrastructure, awaiting activation.
(Unverified leaks; treat timelines, pricing, and benchmarks with caution.)
1
48
Ciprian Nastase retweeted
Feb 2
This is a good thread, and a useful one...
...but you should not assume that the Claude Code team has discovered the actual best ways to use Claude Code. A lesson of the past three years is that figuring out the best way to use AI is an exploratory process that is open to many!
Jan 31
I'm Boris and I created Claude Code. I wanted to quickly share a few tips for using Claude Code, sourced directly from the Claude Code team. The way the team uses Claude is different than how I use it. Remember: there is no one right way to use Claude Code -- everyones' setup is different. You should experiment to see what works for you!
34
9
429
47,987
not exactly coding aisecondbrain.io/p/5-lessons… via @NorthstarBrain
20
How's this for a spiky point of view?
Jan 29
Generally I find people who are anti-AI because they think AI is just hype respond rather well to being shown how AI works in a non-judgmental way. And if they hate AI for other moral reasons, they may still dislike AI after but they become more informed critics & we need those.
1
26
Ciprian Nastase retweeted
Jan 29
Generally I find people who are anti-AI because they think AI is just hype respond rather well to being shown how AI works in a non-judgmental way. And if they hate AI for other moral reasons, they may still dislike AI after but they become more informed critics & we need those.
do you think they’d be less anti AI if they knew about the true capabilities? or just madder
33
13
297
35,398