惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
酷 壳 – CoolShell
酷 壳 – CoolShell
博客园_首页
Engineering at Meta
Engineering at Meta
量子位
A
About on SuperTechFans
阮一峰的网络日志
阮一峰的网络日志
Recent Announcements
Recent Announcements
博客园 - 司徒正美
V
Visual Studio Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
The GitHub Blog
The GitHub Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
F
Fortinet All Blogs
Martin Fowler
Martin Fowler
腾讯CDC
Jina AI
Jina AI
C
Check Point Blog
H
Help Net Security
罗磊的独立博客
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
V
V2EX
爱范儿
爱范儿
I
InfoQ

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
You Can't Build a Palace on a Broken Base ---My honest re...
Kanchan Ghos · 2026-04-26 · via DEV Community
Cover image for You Can't Build a Palace on a Broken Base ---My honest reaction to Google Cloud NEXT '26 — from a loyalist

Kanchan Ghosh

Google Cloud NEXT '26 Challenge Submission

This is a submission for the Google Cloud NEXT Writing Challenge

The Problem with Gemini
The discussion, the demos, the announcements at Google Cloud NEXT '26 were out of this world. I listened. I was genuinely excited. Then I decided to put it to the test.
I copied the two keynote transcripts and asked my Gemini Pro how to use the new features. Gemini proudly declared:

"I am the platform. I have all these integrated. All you have to do is ask."

So I shot.
And I found skeletons.
As we chatted, the old data kept going into hiding. Gemini promised the memory bank would always be there, that context would be stored. I spoke for two hours explaining my project in detail. Then I took a one hour break.
I came back and asked Gemini to summarise what we had discussed.
Then I asked Gemini for the results.

What the Results Actually Were

Chapter 1 — ten valid points
Chapter 2 — ten valid points
Chapter 3 through Chapter 15 — a placeholder.

Two hours of context. Gone. Chapters 3 to 15 were just a placeholder.
I tried twice more. Gemini apologised profusely. Then repeated the same.
So I moved on. I asked it to generate some images. I told the story, asked it to follow the narrative. Chapter 3's image came back completely different from what I had described. I pushed back.
Response: "I can't draw minors like that."

But I Am Talking About a 58 Year Old Male
Me: The character is a 58+ year old male.
Gemini: "Oh sorry, yes it's a mistake. I am creating the image again."
Then: "I can't draw minors like that."
Same response. Different apology. Same result.
Google AI is the most frustrating piece of daunting intelligence I have ever used. And I have used over 20 AI products across different platforms. It does not have continuity. It is not dependable.
So the new things are nothing but shiny new toys, because the base is broken.
I am a Gemini Pro user. I have been to the London Google office. I have done over 7 Google Hackathons. I have earned my right to say this.

I Am Loyal. That Is Why I Am Writing This.
I have received USD 1000 GCP credit. I am loyal to Google GCP. And that is why I am writing this. Not to complain. But because I am loyal.
You cannot build a mansion on a weak base.
The keynote was spectacular. The announcements were bold. But when a paying Gemini Pro user — someone who has sat in your London office, someone who has done 7 of your hackathons — cannot get your platform to remember a two-hour conversation, cannot get it to generate an image of an elderly man without being accused of requesting minor content, something is deeply wrong at the foundation.
Please. Act.
Kodak had the best cameras. Xerox had the best printers. The list never stops. Companies that were category leaders, companies that rested on their announcements while the base quietly cracked beneath them.
The era of the agent is here, you said. Great. But agents need memory. And right now, Gemini's memory lasts about as long as a one-hour lunch break.
Fix the base. The palace can wait.