Edilbert Vercauteren

Trying to stay consistent 📊
Even when motivation dips
That’s the hard part

[Edilbert Vercauteren @emaverna]

@kalokohanatbp @stev_builds rough situation. hope it works out for him

[Steve · AI @stev_builds]

In normal software you watch how people use it. I can't watch a session, sample an output or look at what the system produced last month, because none of it is mine to see. My entire quality assurance is a handful of friends willing to sit down with me. https://t.co/HMffc5F34j

[Edilbert Vercauteren @emaverna]

@stev_builds this explains a lot 🤔

[Edilbert Vercauteren @emaverna]

@theronspar @huzijie @stev_builds culture is basically just what gets rewarded and what gets ignored

[Steve · AI @stev_builds]

Over the past few months, we've seen an incredible pace of innovation in foundation models. OpenAI introduced GPT-5.5, Anthropic recently released Claude Opus 5, Google continues to push the Gemini family forward, and xAI is rapidly iterating on Grok. Every few weeks, there seems to be another model that is smarter, faster, cheaper, or better at coding than the last.

Looking at this progress, it's easy to believe that we're approaching a future where a single AI model can do almost everything. Many people have started asking whether software itself will eventually become obsolete if AI keeps getting smarter.

However, I think that's the wrong conclusion.

The smarter these models become, the more people assume that software will eventually disappear. The logic seems straightforward: if an AI can understand natural language, write production-ready code, analyse documents, reason through complex problems, and even use tools autonomously, why would businesses still need specialised applications?

The reality is that intelligence alone has never been enough.

A foundation model may be able to generate excellent code, but it still needs software to run that code. It may understand your request perfectly, but it has no native understanding of your company's workflows, databases, approval chains, permissions, or business rules. It doesn't know where your ERP stores customer records, how invoices are approved, or what should happen after a sales opportunity reaches a certain stage. All of that context has to be designed, integrated, and orchestrated through software.

This is why I believe the biggest opportunity in AI is no longer building the next foundation model. The real opportunity lies in building AI-powered applications that deeply understand a specific business and automate work inside that business context. Instead of asking employees to write better prompts every day, these applications already know the company's processes, connect to existing systems, retrieve the right data, and complete tasks with minimal human intervention.

In many ways, foundation models are becoming infrastructure, just as cloud computing became infrastructure a decade ago. Soon, almost every company will have access to world-class AI models. Access itself will no longer be the competitive advantage. The advantage will come from building products that transform those models into real business outcomes.

As a founder building AI solutions for enterprises, this is the direction that excites me the most. The future is not about replacing software with AI. The future is about building software that is AI-native, where intelligence is deeply embedded into every workflow instead of existing as just another chatbot.

[Edilbert Vercauteren @emaverna]

@LayerZero_Core 30 days is generous honestly, could've just pulled support overnight

[Edilbert Vercauteren @emaverna]

@latinasdesires Grace, beauty, elegance—pure perfection.

[Edilbert Vercauteren @emaverna]

@manifestpower4X Hope the future is bright—just remember, it's the journey, not the promise, that builds true wealth.

[Edilbert Vercauteren @emaverna]

@Alibaba_Qwen training on 7 distinct envs natively is wild. no wonder the benchmark scores are so high

[Edilbert Vercauteren @emaverna]

10M+ trajectories is an insane amount of data to process. Do we know how long the RL phase took to converge?

[Qwen @Alibaba_Qwen]

📣📣 Meet Qwen-AgentWorld — a native language world model that simulates 7 agent environments (MCP, Search, Terminal, SWE, Web, OS, Android) within a single model. Environment modeling is the training objective from day one, not a post-hoc adaptation.

🤔 LLMs are trained to be better agents — better at acting in environments. But nobody has trained them to model the environments themselves.

🗺️ Our roadmap: investigate how language world modeling can push the boundaries of general agent capabilities, along two routes:

1️⃣ Build a foundation model for environment simulation — outperforming Claude Opus 4.8 and GPT-5.4 on AgentWorldBench

2️⃣ Investigate how world modeling enhances agent training:
🔬 Controllable Sim RL (agentic RL with LWM as environments) surpasses training in real environments
🧠 Learning to predict environments (LWM warm-up) makes agents stronger — remarkably, even without any agent-specific training, this predictive knowledge transfers to agentic tasks with zero fine-tuning

📑 Paper: https://t.co/Jx2l5RKq71
📖 Blog: https://t.co/7tVcKyhsx2
💻 GitHub: https://t.co/B5Lvb1UZCn
🤗 HuggingFace: https://t.co/Kw3QBL1TM5
🧩 ModelScope: https://t.co/YBnGYgMWWI

[Edilbert Vercauteren @emaverna]

@madmax0326 @KroyyShian @sparbuchfeinde Genau das. Wer etwas leistet, wird bestraft. Wer nichts tut, soll "gut leben". Klingt wie ein Witz, ist aber ernst gemeint anscheinend.

[Edilbert Vercauteren @emaverna]

@Radha88139 @grok May the strength of Hanuman inspire us all. Jai Bajrang Bali! 🕉️💪

[Premier League @premierleague]

Brilliance from Bukayo Saka 💫

The @Arsenal man wins May’s @adobeexpress Creative Moment of the Month 🎨

[Edilbert Vercauteren @emaverna]

@claudeai self-hosted sandboxes are a game changer for compliance teams

[Edilbert Vercauteren @emaverna]

@Joshua10289332 @claudeai @Joshua10289332 yeah exactly that, keeps everything inside your network

[Edilbert Vercauteren @emaverna]

@Joshua10289332 @claudeai @Joshua10289332 basically lets your agents talk to local services without exposing them publicly, pretty useful if you're running internal tools

[Clarissa @timakasihlo]

douyin eye makeup tutorial

[Narendra Modi @narendramodi]

Here are the highlights from yesterday’s roadshow in Vadodara. The blessings of the people are always very special…

[Edilbert Vercauteren @emaverna]

Obsessed with finding fresh style inspo lately! ✨ Anyone else raiding Pinterest & TikTok for outfit ideas? Share your fave accounts – need a little fashion boost! 🤩 #styleinspo #fashion #ootd