What it actually feels like to be using AI agents to build an app. pic.twitter.com/ewx1XMYTnL
— Bojan Tunguz (@tunguz) 19 janvier 2026
What it actually feels like to be using AI agents to build an app.
AGENTS
-
The Real Feel of Using AI Agents to Build Apps
By
–
-
Run SWE agents on remote VMs via Agents API
By
–
Now you can run many SWE agents like Claude Code, Codex CLI, Gemini CLI and others on remote VMs, managed with a single Agents API released by @blackboxai
— 🚨 AI News | TestingCatalog (@testingcatalog) 19 janvier 2026
AI Agents outsource 🤖 https://t.co/MMUN4SPlwC pic.twitter.com/I0k7v6BDriNow you can run many SWE agents like Claude Code, Codex CLI, Gemini CLI and others on remote VMs, managed with a single Agents API released by @blackboxai AI Agents outsource
-
Alignment Research Discovers Framing Dependence in Strategic AI Responses
By
–
Further probing has found framing dependence — which, to be clear, you'd not like to see in a default-aligned, universally convergent strategic reply — and not all suggested frame dependence has panned out. But still, framing dependence.
-
ChatGPT 5.2 Auto Responses and Decision Theory Discussion
By
–
Responses by ChatGPT 5.2 Auto. Currently popular prompts found by me at https://
x.com/uubzu/status/2
012782068829171955
… and https://
x.com/gmltony/status
/2012936406461456411
…. Similar result pair at https://
x.com/hovormijurko/s
tatus/2012993464464511270
… with real picture. Longer explanation at https://
lesswrong.com/posts/rP66bz34
crvDudzcJ/decision-theory-does-not-imply-that-we-get-to-have-nice
…. -
Abacus AI: Orchestrating 40 AI Models for Maximum Performance
By
–
Abacus AI Is The Infrastructure You Need To Orchestrate Across 40 AI Models Reasoning – GPT 5.5 thinking and pro
Coding – Opus 4.5 Spatial Intelligence – Gemini 3.0 Pro – Kling 2.6 Motion Control
Images – Nano Banana Pro Routel LLM will max and match the BEST AI Models -
Top AI Research Papers: RAG, Agents, and Memory Optimization
By
–
The Top AI Papers of the Week (January 12-18) – UniversalRAG
– Agent-as-a-Judge
– Self-Evolving Search Agents
– Active Context Compression
– Efficient Lifelong Memory for LLM Agents
– Extending Context by Dropping Positional Embeddings
– Unified Long-Term and Short-Term Memory -
AI agents timeline: MCP to machine-to-machine economy by 2030
By
–
2025: Agents got a shared language (MCP)
2026: Agents learned to talk to each other (A2A)
2027: Agents got wallets
2028: Agents started hiring other agents By 2030, most of the economy runs machine-to-machine. Humans won’t be replaced.
They just won’t be required. -
LLMs Replace Hand-Crafted Rewards in Multi-Agent RL
By
–
A huge claim from this paper on the end of reward engineering. Reward engineering remains a persistent bottleneck in multi-agent RL. This paper argues that LLMs enable a fundamental shift: from hand-crafted reward functions to natural language objectives. If language can
-
After using 15 AI tools, created one that does everything
By
–
I was using more than 15 AI tools… It wasn't possible anymore So, I created my own. And it does EVERYTHING https://
youtu.be/N0gTIxg9l2A -
Deep Agent Projects Coming Soon to Abacus AI
By
–
We already support projects but we will have projects with Deep Agent in a couple of weeks