Deep in Agents, LLMs and automation · Open source at ♥ · Sharing AI gems daily to help you work smarter · Ex-data scientist @Snowflake | @Streamlit ·
📩 DM for Collabs
Joined January 2009
- Tweets35.5K
- Following49.4K
- Followers182K
- Likes11K
Pinned Tweet
YOUR AGENT FINISHED THE TASK. BUT DO YOU KNOW HOW?
The final answer looks fine.
Somewhere along the way, it may have looped on the same tool, re-read the same file, or retried a failed call until it worked.
You only see the bill.
@RequestyAI's Traces view rebuilds the whole agent session, step by step:
→ every model call
→ every tool call, with its arguments and output
→ the cost and latency of each step
The interesting part?
You don't send any extra metadata. The session is rebuilt from the requests themselves.
It won't fix a bad agent for you.
But you can't debug what you can't see 👀
Link in my first reply ↓
🤝 Paid partnership
Charly Wargnier ♨️ retweeted
YOUR AGENT FINISHED THE TASK. BUT DO YOU KNOW HOW?
The final answer looks fine.
Somewhere along the way, it may have looped on the same tool, re-read the same file, or retried a failed call until it worked.
You only see the bill.
@RequestyAI's Traces view rebuilds the whole agent session, step by step:
→ every model call
→ every tool call, with its arguments and output
→ the cost and latency of each step
The interesting part?
You don't send any extra metadata. The session is rebuilt from the requests themselves.
It won't fix a bad agent for you.
But you can't debug what you can't see 👀
Link in my first reply ↓
🤝 Paid partnership
Charly Wargnier ♨️ retweeted
i don't think people realize how big this is…
A US lab just released a 40B open-weight coding model leading its class on three major coding benchmarks.
125+ tok/s on a MacBook. 150+ on RTX Spark.
no per-token fees. no usage limits.
your laptop is now the endpoint.
🤝 Paid partnership
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Charly Wargnier ♨️ retweeted
YOUR AGENT JUST GOT ITS OWN SPACE
@ego_agent lite lets Claude Code, Codex and Cursor work in a separate Space with your existing logins.
you keep your tabs. it does the task.
free. macOS and now Windows.
🤝 Paid partnership
YOUR AGENT JUST GOT ITS OWN SPACE
@ego_agent lite lets Claude Code, Codex and Cursor work in a separate Space with your existing logins.
you keep your tabs. it does the task.
free. macOS and now Windows.
🤝 Paid partnership
Charly Wargnier ♨️ retweeted
most startups celebrate one good week.
Pancake.ai just strung together ten of them, at 50% growth each.
that's not luck anymore.
🤝 Paid partnership
it's been 10 weeks of 50% week-over-week growth with pancake.ai
no matter how many startups you’ve built, the next one still makes you feel like a beginner. there’s always something you’ve never done before that you suddenly need to figure out.
jumping out of a plane, trying to break a guinness world record. it's all fun!
but the best part is still realizing you’ve built something other people think is worth paying for.
more to come!
most startups celebrate one good week.
Pancake.ai just strung together ten of them, at 50% growth each.
that's not luck anymore.
🤝 Paid partnership
it's been 10 weeks of 50% week-over-week growth with pancake.ai
no matter how many startups you’ve built, the next one still makes you feel like a beginner. there’s always something you’ve never done before that you suddenly need to figure out.
jumping out of a plane, trying to break a guinness world record. it's all fun!
but the best part is still realizing you’ve built something other people think is worth paying for.
more to come!
Charly Wargnier ♨️ retweeted
YOUR AGENT CAN RUN ALL NIGHT. YOUR BUDGET WILL NOT.
43% of enterprises struggle to prove AI's business value, while 27% cannot even track token spend across systems.
@StackletIO just launched Token Custodian to close that gap.
An agent stuck in a retry loop can balloon its context window on every turn.
By the time a monthly billing alert arrives, the money is already gone...
Built by the team behind CNCF Cloud Custodian, Stacklet already governs $10B+ in enterprise cloud spend, cutting cloud and AI bills by up to 50% and governance overhead by 80%.
Token Custodian brings real-time enforcement to LLM tokens:
> attribute tokens and agent runs to teams, projects, apps, and cost centers across providers
> cap agent steps, runtime, and context growth per run
> pause runaway loops automatically and alert owners in Slack
> switch routine requests to cheaper models at 80% budget usage, with approvals for frontier model exceptions
FinOps sets budgets in the UI.
Platform teams manage policies as code in Git across coding assistants and customer-facing agents.
The results?
savings 6x faster, with flat subscription pricing instead of a percentage of spend 🔥
now in early preview ahead of the October 27 Token Governance Summit (event link in 🧵↓)
Explore it here → fandf.co/47ns9ta
🤝 Paid partnership
i don't think people realize how big this is…
A US lab just released a 40B open-weight coding model leading its class on three major coding benchmarks.
125+ tok/s on a MacBook. 150+ on RTX Spark.
no per-token fees. no usage limits.
your laptop is now the endpoint.
🤝 Paid partnership
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
YOUR AGENT CAN RUN ALL NIGHT. YOUR BUDGET WILL NOT.
43% of enterprises struggle to prove AI's business value, while 27% cannot even track token spend across systems.
@StackletIO just launched Token Custodian to close that gap.
An agent stuck in a retry loop can balloon its context window on every turn.
By the time a monthly billing alert arrives, the money is already gone...
Built by the team behind CNCF Cloud Custodian, Stacklet already governs $10B+ in enterprise cloud spend, cutting cloud and AI bills by up to 50% and governance overhead by 80%.
Token Custodian brings real-time enforcement to LLM tokens:
> attribute tokens and agent runs to teams, projects, apps, and cost centers across providers
> cap agent steps, runtime, and context growth per run
> pause runaway loops automatically and alert owners in Slack
> switch routine requests to cheaper models at 80% budget usage, with approvals for frontier model exceptions
FinOps sets budgets in the UI.
Platform teams manage policies as code in Git across coding assistants and customer-facing agents.
The results?
savings 6x faster, with flat subscription pricing instead of a percentage of spend 🔥
now in early preview ahead of the October 27 Token Governance Summit (event link in 🧵↓)
Explore it here → fandf.co/47ns9ta
🤝 Paid partnership
BONUS!
Stacklet is hosting the Token Governance Summit on October 27
90 minutes on AI token costs, smarter model routing, and stopping runaway agents, featuring FinOps leaders and AI builders.
Includes a first look at Token Custodian.
Register here → fandf.co/4BjGpko
Charly Wargnier ♨️ retweeted
Open-source models need more than open-weights.
they need fast inference.
this @general_compute and @positron_ai partnership is another step toward putting that infrastructure in builders' hands:
> $15M seed raised
> Reports 30 MW+ capacity
> Cerebras, SambaNova, d-Matrix, Positron
> Bare metal or one endpoint
Positron rollout planned for Q1 2027 👀
🤝 Paid partnership
Our heterogeneous cloud just got better. We're announcing a cluster scale deployment of @positron_ai as our next partner.
We believe demand for fast inference will outstrip supply for the next decade. Our bet on Positron is another step towards getting fast inference for opensource models into builders' hands.
Hardware goes live Q1 2027.
Thanks to @mitesh711, @trsohmers and the whole Positron team! We're excited for this partnership.
LFG!
this is f*cking insane
built the knowledge systems at Stripe and Uber.
helped Anthropic, Metronome and Graphite build their developer experience.
now Dave's launching his own: Falconer.
Falconer can:
> flag stale, missing, duplicate and conflicting docs
> give cited answers across docs, code, tasks and conversations
> turn code, Slack threads, meetings and agent sessions into docs
> update public docs when your product changes
> serve it all to agents over MCP
backed by GP, BoxGroup and SV Angel, with angels from Anthropic, Stripe, Meta, Cursor and Replit...
that should tell you something.
Trillions are pouring into the race to build the first general superintelligence, but leveraging your own specialized know-it-all is now possible as of today.
I built the knowledge systems at @stripe and @uber. Then I helped Anthropic, Metronome, and Graphite build a killer developer experience. Everyone wants Stripe-level quality, but that bar is too low.
Falconer gives you and your customers complete, specialized intelligence: an internal knowledge base and external documentation in one beautiful package. Falconer is the ultimate trust engine to manage your context: self-driving, observable, and agent-friendly.
I co-founded @falconer_ai with @mbenedetto212 and assembled a team of know-it-all builders. We're backed by fellow know-it-alls @thegp, @BoxGroup, @svangel, @karimatiyeh, @benln, @zachlloydtweets, @CeciStalls, @MerrillLutsky, @scottmac, @malthe, and @adamjstevenson, with angels from Anthropic, Stripe, Meta, Cursor, Replit, and more.
Frontier intelligence companies like @FireworksAI_HQ and @⬛⬛⬛⬛⬛⬛; fintech companies like @payabli and @vori; and healthtech companies like @thatch_ai use @falconer_ai because they all share the same vision as us: the knowledge you earn should belong to you.
Now you can know everything, so you can build anything.
I listed my Claude skill on Capafy, and it’s made over $289K in sales in the past 30 days!
It’s called CloneCut. It helps you create your own version of a viral video you like, using your photos and ideas.
The first version came together pretty quickly. Then I started posting videos on Instagram and TikTok to show people what it could do.
Those videos helped me attract my first paying users. Many of them tried it once, then kept coming back and paying for more.
Seeing views turn into actual sales showed me just how effective social media videos can be.
The dashboard is now at $289,462.51.
Thank you to everyone who’s tried CloneCut, shared it, or supported me along the way.
Taking a moment to celebrate this one. Then back to making CloneCut even better.
Try CloneCut on Capafy:
capafy.ai/agent/clonecut-vir…
