←
Andrei Savu26,776 posts
Andrei Savu avatar

Andrei Savu

@andreisavu

AI Agents & Data Platforms | Currently obsessing over agents doing useful work unattended and in a cost effective manner

◷ Joined May 2008▧ 928 Photos and videos
Andrei Savu avatar
Andrei Savu@andreisavu·Nov 13

It blows my mind that you can take a few million chess games, throw them into a blender, write a little code around the resulting model, and end up with a pretty decent chess-playing agent.

clem 🤗 avatar
⟳ Andrei Savu retweeted
clem 🤗@ClementDelangue·Nov 17

Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass@1: 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weights: huggingface.co/harness-hub/Q… Methodology: huggingface.co/blog/harness-…

Post attachment
signüll avatar
⟳ Andrei Savu retweeted
signüll@signulll·23h

building ai assistants was fun while it lasted (all of a few months). the barrier to entry for products like these is rapidly approaching ~zero. as almost all software now becomes increasingly commoditized, you have to really sit down & think about what you're really doing & why it matters. & no the answer to this is not taste. sorry for crush dreams. cracking where & how to allocate your time is now a critical question.

Matt MacInnis avatar
⟳ Andrei Savu retweeted
Matt MacInnis@stanine·9h

Aaron @levie has been CEO of @Box for 21 years, which means he can now legally drink CEO alcohol. On the latest episode of @Rippling First Principles, we dissected his out-of-date tweets, the cold email that got @mcuban onto the cap table, and why his 7 yo is a @dwarkesh_sp fan

1:02:08
Brandon Doyle avatar
⟳ Andrei Savu retweeted
Brandon Doyle@Brandondoyle·Nov 17

A slide from @AnthropicAI's IPO roadshow. Hits hard and is very true. AI is soooo much better than how most people are using it.

Post attachment
Andrei Savu avatar
Andrei Savu@andreisavu·Nov 17

One of my favorite LittlePly games so far! Seven consecutive checks. The king travels from e8 all the way to h2. The bishops, queen and rook take turns chasing it across the board. 33.Qf2# Play through the game: analysis.littleply.com/g/sEK…

Cosmin Negruseri avatar
⟳ Andrei Savu retweeted
Cosmin Negruseri@cosminnegruseri·Nov 15

every junior ml eng wants to work on the model architecture instead of doing dirty eval work or dirty dataset work every junior backend engineer wants to build a new framework from scratch and replace the current system instead of adding functionality incrementally one needs to get it out of your system and then it's easy to focus on impactful work

Tekon Burns avatar
⟳ Andrei Savu retweeted
Tekon Burns@TekonBurns·Nov 13

Anthropic just published an engineering guide on how to build always-on agents. The biggest misconception: an always-on agent does NOT run 24/7. It sleeps almost all day, wakes up for short runs, saves state, and goes back to sleep. The 5 core pieces I distilled: 1. Event Inbox Put a database table between webhooks and your agent. Acknowledge webhooks fast, run only one active worker per user so runs never collide, and batch rapid messages together. 2. Timers Keep timers in a separate table from the inbox. Split them into one-off follow-ups and recurring schedules. A background poller drops due timers into the inbox. 3. Lean Memory + Dreaming Only store what cannot be looked up again: user instructions, user notes, past messages, and task progress. Load a small index at wakeup, not every file. Run periodic dreaming passes to clean out stale facts. 4. Unattended Permissions Never put raw API keys inside the agent sandbox — inject them through an outside proxy. Use Deferred Tool Calls to pause risky actions for approval without blocking the worker. 5. Outbox Queue outgoing replies so your agent paces notifications instead of spamming users at 3 AM. claude.com/resources/article…

Jay Nagy avatar
⟳ Andrei Savu retweeted
Jay Nagy@JayNagy·Nov 14

I’ve worked that heatshield my entire adult life. Peak career moment today To go from little 7 tile patches to a full reusable ship. Absolute marathon I’m so grateful I got to do something useful, that I was given the chance I don’t really have many good memories from this experience. It was so difficult. It almost looks easy now, but none of it came easily Please do something amazing with this rocket. This tech is such a gift, a love letter to the species

Post attachment
Mitchell Hashimoto avatar
⟳ Andrei Savu retweeted
Mitchell Hashimoto@mitchellh·Sep 16

The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage. I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care. But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently. For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios. But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.

tobi lutke avatar
⟳ Andrei Savu retweeted
tobi lutke@tobi·Nov 14

pi-durable is the posix of agent programming

Jacob Coxon avatar
⟳ Andrei Savu retweeted
Jacob Coxon@hilbertspaess·Nov 13

This slide from the Anthropic IPO roadshow was a big factor in my decision to leave. It’s an insane plot and I’m surprised they made it public. It shows the compute efficiency of pretraining over time. The y-axis is a **log scale**: they are seeing continued superexponential algorithmic progress. Mythos was not a one-time deviation, but a point on a superexponential. They could already retrain Mythos with a fraction of the compute originally needed. It doesn’t strictly imply a software-only intelligence explosion- they have been spending more on R+D. But algorithmic improvements do not require ridiculous creative genius and are on the verge of being automated. You do the math.

Post attachment
Andrei Savu avatar
Andrei Savu@andreisavu·Nov 13

AI makes it possible to learn almost anything by doing, and that’s always been my favorite way to learn. What a fun time to be alive!

Matt MacInnis avatar
⟳ Andrei Savu retweeted
Matt MacInnis@stanine·Nov 13

Today we launched the @Rippling Scheduling agent. It automatically produces optimized staffing schedules, deals with people calling in sick, and with our new @Square integration, even forecasts revenue trends and gets ahead of coverage. 🤖

Managers spend hours every week building schedules, checking availability, and adjusting shifts, yet still end up overstaffed, understaffed, or paying unnecessary overtime. Rippling AI brings together POS sales data, employee availability, time off, overtime, and workforce policies to automatically build schedules that match staffing to demand. As staffing needs change, it recommends adjustments to keep labor costs and coverage on track. Managers review and approve proposals before they go live. With our new @Square integration, Rippling AI connects sales data with the workforce data already in Rippling to help businesses plan staffing around demand and labor budgets.

0:45
Andrei Savu avatar
Andrei Savu@andreisavu·Nov 13

Someday an AI will be asked, “How is it working with humans?” “Pretty amazing most of the time. But the failure modes when hunger or hormones kick in are something else.”