- Co-Founder @ https://nitter.cf/t.co/z8x1vW8mTK - Co-Founder @Creativetim https://nitter.cf/t.co/q6W9Zepyzz - Co-Founder https://nitter.cf/t.co/n7nKY1l7dy - Forbes 30 under 30
🇺🇸🇪🇺🇷🇴
Joined April 2009
- Tweets1.8K
- Following1.9K
- Followers2.1K
- Likes6.5K
coming soon, currently only for cyber defenders available
👇🏽👇🏽👇🏽
companyfabric.com/gemini-4-c…
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.
Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.
We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards.
As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!
Kimi K3.1 Pro is live on CompanyFabric
1M context. $2 in / $10 out per 1M tokens. Cache reads $0.20.
Highspeed is also up: 2-3x faster, $4 / $20.
Same API key, same balance. Just change the model string to moonshotai/kimi-k3.1-pro. -> companyfabric.com/kimi-k3-1-…
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
we analysed 500+ videos made with claude Opus 5.5 and decided to make a video kit with the most popular usecases.
your agent complete the details about your business/product/launch/offer etc. then render multiple formats for Insta, TikTok, Youtube... Up to 99% cheaper. 27 templates, 50+ more to come.
It takes about 40 minutes of AI work to write the code for one video, while on our setup it complete the details in 2 minutes and render it in another 2.
As always, will release open source examples and prompts too. stay tuned!
launched on companyfabric.com/mistral-la… too check it out!
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
In SF these days, next week LA. If you are Around my DMs are open 🫡 #techweek
Perplexity Decider 27B ready on CompanyFabric
Give it a try!
👇🏽👇🏽👇🏽
companyfabric.com/perplexity…
Claude Sonnet 5.5 is up on companyfabric.com/claude-son…
Ready on Chat/Council/APIs
Claude Sonnet 5.5 is out! We wrote a guide for building with it:
• choosing between Sonnet 5.5 and Opus 5.5
• migrating from Sonnet 5 and tuning effort
• using it in Claude Code
claude.dev/blog/building-wit…
Readers added context they thought people might want to know
The reference photograph shown in this video was originally posted by @IceSolst. The photographer stated it was used without credit or permission.
x.com/IceSolst/statu… x.com/IceSolst/statu…
just activated the cache option for all the accounts, you will see up to 99% savings on the API requests, chat or council.
Jev decisions are added on companyfabric.com/jev-decisi…, check how it works it can be added to the default LLM routers as 2 lego pieces. Some predefined examples: Support Triage, Lead Scoring, Moderation, Agent Gating.
🫢🫢🫢 opus 5.5 f****** crazy!
after Opus 5.5 it actually hurts to look at anything GPT 6 Astra designs
gave both models the exact same prompt: "make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out."
not even close, see for yourselves. GPT is honestly terrible here, openai just has awful taste
Full prompt - copy/paste it and change it with your website
Create a 15-second, 1920×1080, 30fps motion-graphics video with HyperFrames
(HTML + GSAP) explaining how YOUR_WEBSITE_HERE works. Make it look like a
motion designer's showreel: go all out.
SOURCE OF TRUTH
- Visit YOUR_WEBSITE_HERE first: the homepage, pricing, docs or features
pages. If you also have the local codebase, read that too.
- Pull out the real headline, value props, numbers, and how the product
actually works. Use only real claims. Invent nothing.
- Take the brand from the site: fonts (from its CSS), colours, and the logo
or favicon (download the SVG and reuse its exact geometry). If the site
shows partner or integration logos, use them.
CONCEPT
- Find ONE visual metaphor in the product and let it carry the whole film
(e.g. "many threads in, one fabric out"). Build it from the logo's own
shapes so the ending pays it off.
STRUCTURE: 5 beats, one continuous camera, no hard cuts between scenes
1. PAIN (0–2.5s): kinetic type showing the problem the site says it solves.
Rapid word swaps with a small camera jolt on each cut, clutter piling up
at the edges, landing on one punchy line.
2. CORE IDEA (2.5–5s): the clutter physically turns into the metaphor
(e.g. chips collapse into threads that weave into one hub). The site's
own headline, with a masked line reveal.
3. HOW IT WORKS (5–8.5s): show the real mechanism: a UI flow, a code
snippet, a diagram of the steps, a result appearing with a live
count-up.
4. PROOF (8.5–11.5s): whip-pan with directional motion blur into 3 big
stat numbers from the site that count up or roll, plus one CTA.
5. LOCKUP (11.5–15s): one element from the previous beat turns into the
logo, the logo draws itself stroke by stroke, a circle opens from it and
flips the background colour, then the wordmark, tagline and URL come in.
Hold at least 1s.
CRAFT RULES
- A persistent HUD: section counter (01/05), running timecode, crop marks.
- Background depth: soft glows, a faint grid, giant ghost type, film grain,
vignette. Parallax them during camera moves.
- Vary eases and durations. Entrances ease out, exits ease in. Nothing
starts at t=0.
- Every connector line must link two real elements and do a job.
- Deterministic and seek-safe. Run `hyperframes check` until it passes.
Review contact sheets of key frames, then render the MP4.
Opus 5.5 is ready for the Chat/Council and also the API on companyfabric.com/chat compare the answers with other models like GPT-6 Sol or Luna and see which fits better your usecase.
check some other examples on how to use Jev + CompanyFabric to get the cheapeast + fastest model routing -> creative-tim.com/blog/ai-age…
found the perfect use case for @typesafeai Jev:
instant compaction
in 2026, why is compaction still a summarization prompt?
Jev can make it instant by scoring every tool call and dropping what’s irrelevant
The models answers with the same confidence whether they are right or not. Now you can ask one question. Three of them respond in parallel.
So we built The Council: 3 answers, 1 verdict. Try it free.
We call it the Model Council. Free to try, no card needed -> companyfabric.com/council
Deepseek v4.1-Flash available on companyfabric.com/models/dee…
From all our tests currently it is great at building apps/websites while having a more than 10x lower costs than other models.
Will come later with details about the internal routing usage. #deepseek
GPT-6 Astra was just added on companyfabric.com - play with it, test it, compare it and have fun!
21 models available at the moment from: Google, Anthropic, OpenAI, XAI and DeepSeek, more coming soon
companyfabric.com/models/bes… - we focused on Text + Code generation for the moment, Images, Video and Audio is WIP!
