Leo (@LeosReal) header
Leo (@LeosReal) profile picture

building the customer agent @outlit_ai | @ycombinator 🇨🇦 🇸🇻

San Francisco
Joined December 2009
Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)Photo by Leo (@LeosReal)
Superset is now Paradev: The code editor that helps engineers run armies of coding agents. We just raised an $11.5M Seed round led by @usv with @ycombinator to redefine how software is written.
140
74
116
683
882,826
Announcing SWE Odyssey, a standardized benchmark in our family of ultra-long-horizon evaluations. As benchmark after benchmark saturates, we wanted a quantifiable way to test a larger claim: can agents work autonomously for hours and still build the correct thing?
18
21
9
242
40,064
introducing Motion V2, the frontier AI for motion design. this launch video was made entirely with Motion V2. 👇🏽 repost + comment "MOTION" for free credits. tag @motion_so in any post on your X feed for a surprise. here’s how it works + what's changed:
95
47
37
196
145,157
Leo retweeted
Beyond excited to join @NotionHQ where I will be leading Model Research, alongside the whole ZeroEntropy team!
We're excited to announce that @NotionHQ  has acquired ZeroEntropy. We first got to know the Notion team as one of our top customers: Notion AI has been running on our reranker zerank-2, replacing models from larger, better-funded competitors with an 85% latency improvement. Now our whole team is joining Notion to keep building. Alongside the acquisition: 🔓 All of our models are now open source under Apache 2.0 on @huggingface , run them through any inference provider, no commercial license required. 📅 All ZeroEntropy products remain fully supported until September 4th, 2026, after which they will be sunset. Our migration guide covers every path, and our team is available on Slack, Discord, and email to help every customer transition. Full story on our blog and migration guide below. Thank you to our customers, investors, and partners for believing in ZeroEntropy. It's been a privilege to serve you, and we can't wait to keep building things you love at Notion.
44
14
4
243
44,024
Leo retweeted
حياة البرشلوني اذا النهائي بين الارجنتين و اسبانيا
148
4,520
187
37,609
752,365
today, we’re launching a new way to interact with Motion. no more prompting & praying. just talk. start with a simple message like "yo" or "what's trending on tech twitter today?" @motion_so can do research, suggest assets, pull context from @github & @NotionHQ, propose ideas, and refine the creative direction with you. this launch video is made entirely with Motion. 👇🏽 now available at motion [dot] so.
31
19
3
92
20,212
I can’t believe Anthropic comparing their product to nuclear weapons 800 times backfired on them. I am shocked
133
735
46
17,834
512,554
we’re launching Motion, the frontier agent for tasteful motion design. this launch video is made entirely with Motion. 👇🏽 QT + comment "MOTION" to get 1,000 free credits. tag @motion_so in any X post for a surprise. now generally available at motion [dot] so.
260
96
76
594
935,176
1. start by giving Motion a prompt. include any context like product links, X threads, YouTube videos, personal assets, and more. Motion will reference styles, incorporate research, and storyboard. example prompt “Create a launch video for the Teenage Engineering TP-7 recorder. Find assets from teenage.engineering/products… to animate in.”
4
3
1
37
6,677
Mosaic has raised a $3.8M seed round to build video editing agents. What started as a side project to edit our own YouTube videos has quickly turned into something much bigger. Today, global agencies, platforms, and news networks rely on our AI video editing workflows to scale content production. I am especially excited to announce our partnerships with: 1. TubeScience: Meta's largest ad creative partner, 8K videos a month, 100M views a day, $2B annual spend. 2. News Corp: one of the world's largest media organizations, operating businesses like The Sun, TalkSport, Wall Street Journal, MarketWatch, HarperCollins, The Times, and more. We are proud to be backed by top class Silicon Valley investors, like @ycombinator, @MayfieldFund, @ElevCap, @pioneer_fund, @transposevc, @twentytwovc, @scriptcapital, @phosphorcap, @amino_capital, @OliveAssetMgt, @FogVentures, @DCGco, and angels from @ycombinator, @OpenAI, @GoogleDeepMind. Special shout-out to @garrytan, @harjtaggar, and @bosmeny for being our partners through YC and beyond. The funds from this round will be used to add a few more people to our lean team in San Francisco and continue research & development on the frontier of multimodal AI and agentic video editing. I'm incredibly grateful to the entire Mosaic team for all the hard work each and every one of you do. I am so pumped for what is ahead. Follow along our journey @mosaic_so and read more about the announcement at mosaic [dot] so.
80
34
11
366
53,320
introducing superset the IDE for the AI agent era stop waiting on coding agents run them in parallel
125
71
57
1,043
748,258
Leo retweeted
it is insane how far behind opus 4.6 max has fallen behind codex 5.4 xhigh it's not close one is like talking to a coked out intern who hasn't slept in 40 hours taking every shortcut known to man and the other is a prime rigorous mathematician with a German bedside manner
183
54
26
2,094
264,650
Leo retweeted
Today we’re announcing that Respan has raised a $5M seed round led by @GradientVC. We’re building the self-driving observability, evals, and gateway for AI agents. 100+ teams already use Respan. 1B+ logs/month. 2T+ tokens/month. And we’re just getting started.
90
49
12
392
39,738
Someone needs to explain to me why Codex is severely outperforming Claude Code for me. The difference is literally night and day, what did Anthropic do?
1
1
3
402
Excited to finally share Ebla-1 and the C⁴ benchmark. Really enjoyed working with HUD on the evals behind it.
Aviro is introducing Ebla, a state of the art grounded reasoning model. In collaboration with HUD, the Aviro team built C⁴ — a benchmark for long-horizon tasks in corporate document sets. We evaluate four dimensions: Correctness, Completeness, Composition, and Citations. @aviro_ai post-trained GPT-OSS 120b to achieve SOTA performance, with a Pass@1 score of 25.4% and Pass@8 score of 37.1%.
3
3
1
16
2,677
Leo retweeted
we just shipped v1 of our proactive churn signals. the breakdown: assertion → what's happening reasoning → the logic and data behind our conclusion evidence → the truths backing up the analysis timeline → how the situation unfolded across systems next steps → preventive measures signals and customer identities unified across @posthog @stripe @firefliesai @slack and more
2
1
7
764
3 months ago we started building an IDE for the AI agents era Today we’re introducing Superset: an open source app for running any agents in parallel. And today we launched Chat 😉, enabling us to build way more superpowers into how users interact with agents (@ context, link Linear tickets, / commands, etc)
40
20
7
216
60,056
CLI > MCP
6
2
20
2,502