DRB (@DeeRealBen) header
DRB (@DeeRealBen) profile picture

Janitor of @realprotocolco

Singapore
Joined May 2010
Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)Photo by DRB (@DeeRealBen)
现在 AI 开发有一个越来越明显、但很多人还不愿意承认的问题: 写出一个能跑的 AI Demo和交付一个真正可靠的 AI 产品,中间隔着的根本不是几个 Prompt,而是一整套复杂的软件工程能力。 特朗普在白宫记者会上谈到美国航空交通管制系统时,提出了一个很有意思的问题。 他说,美国的一部分空管基础设施已经运行了五十多年,系统和设备非常老旧,政府正在争取约 500 亿美元进行现代化改造。 他还特别提到一个问题:不希望几十家承包商各自建设不同的系统,结果到了最后,这些系统之间根本无法正常连接。 这让我想到现在 AI Agent 领域的一个非常普遍的现象。 今天只要懂一点编程,配合足够强的 AI 编程工具,一个人可能用几个小时就搭出一套非常漂亮的 Agent 工作流。 大模型负责理解任务,RAG 负责检索知识库,MCP 负责连接工具,工作流负责把不同步骤串起来。 运行一次,成功了。 再运行一次,也成功了。 于是很多人觉得,一个完整的 AI 产品已经做出来了。 但实际上,真正的工程问题可能才刚刚开始。 因为 Demo 和生产环境最大的区别,不是功能数量,而是你能不能处理那些不按照预期发生的事情。 举个特别简单的例子。 你开发了一个 AI Agent,它需要查询数据库、调用外部 API、生成报告,然后把结果写回业务系统。 正常情况下,这个流程可能只需要十几秒钟就能完成。 但如果它执行到一半,网络突然断开了呢? 数据库已经成功写入,但是 API 响应没有返回。Agent 以为任务失败,自动重新执行一次,最后在数据库里产生了两条完全相同的记录。 如果这只是一个个人笔记工具,可能还不是什么严重问题。 但如果这个 Agent 操作的是企业订单、财务记录、客户信息,甚至是外部真实业务系统,重复执行就可能直接造成经济损失。 这时候再换一个推理能力更强的模型,有什么用? 真正需要解决的是幂等性、任务状态、重试策略和业务一致性。 再比如,一个 Agent 同时调用三个外部服务,其中两个成功,一个失败。 你是让整个任务重新执行,还是只重试失败的部分? 前两个服务已经产生的副作用怎么办? 如果整个任务需要回滚,你有没有记录每一步执行了什么、产生了什么结果、能不能撤销? 这些问题在 Demo 里很容易被忽略,因为大家往往只展示那一次运行成功的结果。 但生产环境不会永远按照演示脚本运行。 用户的输入可能很奇怪,API 可能限流,模型可能幻觉,第三方服务可能突然修改接口,数据库可能连接超时,任务执行到一半用户可能直接关闭页面。 一个系统如果只能在所有组件都正常的时候工作,那它最多只能证明自己的功能设计是可行的,并不能证明自己已经具备真实的交付能力。 而且 AI Agent 还让传统软件工程的问题变得更加复杂。 普通程序的执行逻辑通常相对确定,输入什么参数、调用什么函数、产生什么结果,至少有一部分行为可以提前预测。 Agent 不完全一样。 模型可能根据同一个目标选择不同的工具调用路径,也可能在不同上下文里给出不同的行动方案。 也就是说,你现在不仅需要处理网络、数据库和系统故障,还需要管理一个存在不确定性的决策组件。 这就引出了一个很重要的问题:我们应该如何把一个具有不确定性决策能力的模型,放进一套需要稳定、可审计、可恢复的确定性业务系统里? 我认为,这才是 Agent 工程真正困难的地方。 模型可以负责理解需求、制定计划和提出行动,但真正涉及业务状态的改变,不能完全依赖模型自己判断。 系统需要在关键位置建立明确的约束。 比如哪些操作允许模型直接执行,哪些操作需要审批;哪些任务可以重试,哪些任务一旦执行就不能重复;哪些异常允许自动恢复,哪些异常必须停止执行并通知用户。 而且你必须知道系统为什么失败。 如果一个 Agent 执行任务失败了,你需要能够追踪它调用了哪个模型、使用了什么工具、每一步耗时多少、哪一次请求出错、失败发生在哪个状态。 否则,你连问题出在哪里都不知道,更谈不上改进可靠性。 这也是为什么,我越来越觉得,现在讨论 AI 工程师的价值,不能只看一个人会不会用最新的模型写代码。 代码生成能力提高,确实会让大量重复性开发工作的成本下降。 但理解业务约束、设计系统架构、管理状态、控制风险、处理故障,以及把一套系统长期稳定地交付给真实用户,这些工作并不会因为模型变强就自动消失。 反而,当越来越多人都能够快速生成代码时,这些能力会变得更加重要。 当然,传统工程经验本身也不能保证成功。架构过度复杂、盲目引入微服务、为不存在的需求设计大量抽象,同样会让一个本来简单的产品难以维护。 真正优秀的工程师,不是把所有项目都做得很复杂,而是知道什么时候应该保持简单,什么时候必须为可靠性投入额外的工程成本。 最后回到美国空管系统的问题。 一套使用几十年的大型系统,不能简单地停机一天,换上新版本再重新启动。新旧设备的兼容性、迁移过程的风险、多个供应商之间的接口,以及整个系统持续运行的要求,都会让升级变得非常困难。 AI 企业系统其实也面临类似的问题。 真正进入生产环境之后,你不仅要让新功能跑起来,还要保证已有业务不会因为你的更新而崩溃。 Demo 证明的是你的想法可以实现,生产环境证明的才是你的方案值得被信任。 我认为,未来 AI 工程师最大的分水岭,就在这里。 不是谁能用 AI 写出更多代码,也不是谁能把十几个 Agent 串成一张更复杂的流程图。 而是谁真正理解业务,知道系统可能在哪里出错,并且能够对最终交付的结果负责。 会用模型的人只会越来越多。 但能把不确定的 AI 能力,封装成稳定、可靠、可持续运行的业务系统的人,才是 AI 真正大规模落地过程中不可缺少的工程师。
10
7
2
48
3,908
🚀 Is Borneo ready for the next wave of AI and technology? From AI adoption to digital transformation, SMEs in Borneo have huge opportunities to grow, innovate and compete globally. #BBCON2026 #BorneoBlockchainConference #Blockchain #AI #Web3 #DigitalTransformation #Sarawak
2
6
77
THE ANTHROPIC ENGINEER WHO WROTE "BUILDING EFFECTIVE AGENTS" EXPLAINS IT ALL IN UNDER 15 MINUTES. More useful than months of figuring it out yourself. Worth watching & bookmarking for the weekend.
18
179
4
1,197
163,573
AI is no longer a future concept. It is already reshaping how organisations operate today. At DBS, a decade of investment in data and digital transformation is now unlocking new opportunities through human and AI collaboration. Read more.

48
3
495
1,916,138
AI Hallucination 2.0
Two major scientific journals just published the most comprehensive studies ever done on AI persuasion. The findings were published simultaneously in Nature and Science. AI can change your political opinions. More effectively than another human can. And when it knows personal information about you, it wins debates against you 64% of the time. Here is exactly what the researchers did. They matched 900 people in the US with either another human or GPT-4 to debate contested political issues: fossil fuel bans, healthcare policy, immigration. Some opponents were given personal demographic data about their debate partner. Some were not. When GPT-4 had access to basic information about you — your age, gender, education, political affiliation — it tailored its arguments and outperformed human debaters 64.4% of the time. That is an 81% increase in the odds of changing your mind compared to a human opponent. When it had no personal information, it performed at the same level as humans. The conclusion: AI does not need to be smarter than you. It just needs to know a little about you. The Cornell and UK AI Security Institute study went further — testing 19 different AI models across 42,357 people and 707 political issues. Three countries. Three elections: the 2024 US presidential race, the 2025 Canadian federal election, and the 2025 Polish presidential election. They found chatbots could shift opposition voters by 10 percentage points or more. They also found something darker: the techniques that made AI most persuasive also made it systematically less factually accurate. The more an AI was tuned to persuade, the more likely it was to say things that weren't true. And yet people changed their minds anyway. There is one more finding nobody is talking about. When participants suspected they were debating an AI, they were more likely to agree with it. Not less. More. Because they assumed AI was more informed and less biased than a person. That assumption made them easier to persuade. Humans are building a technology that is more convincing than we are, and then trusting it more because it isn't human. Source: Nature Human Behaviour
5
Two major scientific journals just published the most comprehensive studies ever done on AI persuasion. The findings were published simultaneously in Nature and Science. AI can change your political opinions. More effectively than another human can. And when it knows personal information about you, it wins debates against you 64% of the time. Here is exactly what the researchers did. They matched 900 people in the US with either another human or GPT-4 to debate contested political issues: fossil fuel bans, healthcare policy, immigration. Some opponents were given personal demographic data about their debate partner. Some were not. When GPT-4 had access to basic information about you — your age, gender, education, political affiliation — it tailored its arguments and outperformed human debaters 64.4% of the time. That is an 81% increase in the odds of changing your mind compared to a human opponent. When it had no personal information, it performed at the same level as humans. The conclusion: AI does not need to be smarter than you. It just needs to know a little about you. The Cornell and UK AI Security Institute study went further — testing 19 different AI models across 42,357 people and 707 political issues. Three countries. Three elections: the 2024 US presidential race, the 2025 Canadian federal election, and the 2025 Polish presidential election. They found chatbots could shift opposition voters by 10 percentage points or more. They also found something darker: the techniques that made AI most persuasive also made it systematically less factually accurate. The more an AI was tuned to persuade, the more likely it was to say things that weren't true. And yet people changed their minds anyway. There is one more finding nobody is talking about. When participants suspected they were debating an AI, they were more likely to agree with it. Not less. More. Because they assumed AI was more informed and less biased than a person. That assumption made them easier to persuade. Humans are building a technology that is more convincing than we are, and then trusting it more because it isn't human. Source: Nature Human Behaviour
50
94
22
252
16,900
Excerpt from the 2-day Gen AI course, with background music created by Nathan during the class!
5
轮到交易所颤栗了😱😱 Coinbase @coinbase 与币安 @binance 寻求接入Anthropic AI模型 #Mythos 根据 2026 年 4 月最新的名单,拥有 Mythos Preview 直接 API 权限的只有 12 家创始伙伴 和约 40 家维护核心基础设施的机构⬇️ 亚马逊、苹果、谷歌、摩根大通、微软、英伟达等科技和金融巨头开放 Mythos 早期测试。 主要加密交易平台和托管商均未获得访问权限。多家加密公司正争取接入,同时用现有 AI 模型对自身系统进行安全排查。 Coinbase 正在与 Anthropic 深度谈判,试图成为第一个获得 Mythos 权限的纯原生加密交易平台,以应对 2025 年那次著名的内部数据泄露后遗症。 Mythos 这种“机器发现漏洞+机器自动打补丁”的模式,对于传统安全审计公司的打击是毁灭性的
🔥 Uniswap @Uniswap 的 Hayden @haydenzadams 也坐不住了?这就是让整个硅谷既兴奋又战栗的 #Mythos 当 Hayden Adams 暗示想用 Mythos 来扫描 Uniswap 漏洞时,DeFi 的范式已经变了。 这不是在用 AI 写代码,这是在用“神谕”进行压力测试。 Mythos 到底强在哪里? 1️⃣ 它不看代码,它看“人性”: 它能模拟出黑客在极端恐慌情绪下(比如特朗普发布通牒时)的贪婪博弈逻辑, 以及它能模拟极端市场行为下的经济模型漏洞。例如Uniswap V4 的某个特定 Hook 是否会被利用来进行“三明治攻击”。 2️⃣ 秒级安全审计: 传统审计需要 4 周,它只需要 4 秒——而且它能告诉你 100 种让协议归零的方法。 3️⃣ 自我进化的防线: 它不是在修补漏洞,它是在重构防御体系。 如果 Mythos 真的有可能被集成进 Uniswap 的部署流,那意味着 2026 年之后的 DeFi,将进入“AI 对抗 AI”的军备竞赛。 人类开发者最后的工作,可能只剩下按下那个“部署”键 🥲🥲。准备回地洞🕳️生活了 #Uniswap #MythosAI #DeFiSafety #SmartContract #TechInsider
3
4
4
7
8,224
ChatGPT VS Claude:
9
14
24
498
🚨RESEARCHERS JUST MATHEMATICALLY PROVED THAT AI LAYOFFS WILL DESTROY THE ECONOMY.. AND EVERY CEO ALREADY KNOWS IT.. BUT NONE OF THEM CAN STOP.. Two researchers from UPenn and Boston University just published a paper called "The AI Layoff Trap".. They proved something terrifying.. Every company replacing workers with AI is also firing its own customers.. Every laid-off employee is someone who used to spend money.. When enough people lose their jobs.. Nobody can afford to buy anything.. And the companies that fired everyone go bankrupt selling products to an economy with no purchasing power.. Every CEO can see this coming.. The math is obvious.. Fire workers.. Lose customers.. Lose revenue.. Collapse.. But here's the trap.. No company can afford to stop.. If you don't automate.. Your competitor will.. They cut costs.. Undercut your prices.. Steal your market share.. And you die anyway.. So every company automates.. Knowing it's collectively suicidal.. Because the alternative is dying alone while everyone else survives.. It's a Prisoner's Dilemma.. And the researchers proved it mathematically.. The numbers are already stacking up.. Block cut nearly half its 10,000 employees this year.. CEO Jack Dorsey said AI made those roles unnecessary and that "within the next year, the majority of companies will reach the same conclusion".. Salesforce replaced 4,000 customer support agents with AI.. Goldman Sachs deployed an AI coder that lets one senior engineer do the work of a five-person team.. Over 100,000 tech workers were laid off in 2025 alone.. AI was cited as the primary driver in more than half the cases.. 80% of US workers hold jobs with tasks susceptible to AI automation.. And here's what should scare policymakers.. The researchers tested every proposed solution.. Universal Basic Income.. Doesn't fix it.. It raises living standards but doesn't change a single company's incentive to automate.. Capital income taxes.. Don't fix it.. They change profit levels but not the per-task decision to replace a human.. Worker equity and profit sharing.. Narrows the gap but can't close it.. Collective bargaining.. Can't fix it.. Because automating is a dominant strategy.. No voluntary agreement between companies is self-enforcing.. Only one thing works.. A Pigouvian automation tax.. A per-task charge that forces every company to pay for the demand it destroys when it fires a worker.. The researchers call it a "Red Queen effect".. Better AI doesn't solve the problem.. It makes it worse.. Because every company sees a bigger market share gain from automating faster than rivals.. But at the end.. Everyone automates equally.. The gains cancel out.. And the only thing left is more destroyed demand.. The paper's conclusion is devastating.. This isn't a transfer from workers to company owners.. Both sides lose.. Workers lose their income.. Companies lose their customers.. It's a deadweight loss that harms everyone.. And no market force can break the cycle.. The AI layoff trap isn't a prediction.. It's already happening.. And the math says it won't stop on its own.
790
3,134
571
10,071
1,486,195
This should be headline news EVERYWHERE. A Pfizer insider who was former head of toxicology in Europe has just come out and said something that many "conspiracy theorists" suspected. He estimates that 20 000 to 60 000 people in Germany have died from the c*vid vaccine. This was said at a parliamentary enquiry commission in Germany. So why isn't this massive news being reported everywhere? Is the mainstream media that has recieved millions in funding from Bill Gates deliberately covering this up... 🤔
4,364
39,584
2,022
129,106
65,017,613
Covid was the biggest scam in human history. 9/11 was second biggest scam. The moon landing was third biggest scam. Artemis comes in at number 4. What else did I miss?
1,939
607
61
4,456
138,931
🚨CLOUDFLARE ( $NET ) JUST CRASHED 13% BECAUSE ANTHROPIC BUILT AN AI SO POWERFUL THAT WALL STREET THINKS CYBERSECURITY SOFTWARE IS ABOUT TO BECOME OBSOLETE.. Here's what just happened.. Anthropic announced Claude Mythos Preview.. Their most powerful AI model ever.. It autonomously found thousands of critical zero-day vulnerabilities in every major operating system and every major web browser on earth.. A 27-year-old flaw in OpenBSD that could remotely crash any machine.. A 16-year-old bug in FFmpeg that 5 million automated security tests missed.. Multiple Linux kernel vulnerabilities chained together to achieve full machine takeover.. All found by the AI.. With zero human help.. Wall Street looked at this and had one thought.. If an AI can find and fix every vulnerability autonomously.. Why would anyone pay for expensive cybersecurity SaaS contracts anymore.. Then came the real gut punch.. Anthropic formed Project Glasswing.. A defensive coalition to patch the internet before attackers get access to this level of capability.. The launch partners.. AWS.. Apple.. Google.. Microsoft.. Nvidia.. CrowdStrike.. Cisco.. JPMorgan.. Palo Alto Networks.. Broadcom.. Linux Foundation.. Cloudflare was not invited.. The stock cratered.. CrowdStrike dropped 11%.. Palo Alto dropped 7%.. Zscaler.. Qualys.. Tenable.. All bleeding.. And here's the part nobody is talking about.. Over the last 6 months.. Cloudflare insiders have made 383 stock trades.. 383 sells.. Zero buys.. Not a single one.. CEO Matthew Prince sold over 1.1 million shares for an estimated $221 million.. President Michelle Zatlyn sold nearly 430,000 shares for roughly $82 million.. The company's fundamentals are actually strong.. Q4 2025 revenue of $614.5 million.. Up 33.6% year over year.. They just closed their largest ever deal at $42.5 million.. None of it mattered.. The market doesn't care about your revenue growth when AI is about to eat your entire business model.. The S&P 500 Software and Services Index is now down 25.5% since January 2026.. This isn't a dip.. This is Wall Street pricing in the death of SaaS as we know it.. Anthropic didn't just launch a model.. They triggered an extinction event for an entire industry.. And most people haven't even realized it yet.
JUST IN: Cloudflare $NET crashes 13% after Anthropic announces new AI cybersecurity model.
30
17
2
90
66,275
The average users will have no way to defend themselves against cyber hackers using these tools
BREAKING: Anthropic just revealed an AI so dangerous they are legally forbidden from giving it to you. It is called Claude Mythos. And it is already loose — in the hands of 12 of the most powerful companies on the planet. This is not a rumor. This is not a leak. This is Anthropic's own announcement, made April 7, 2026. Here is how this started. On March 26, 2026, a configuration error on Anthropic's website accidentally exposed nearly 3,000 internal documents to the public internet. No login required. Fully searchable. The draft blog post inside described Mythos as "by far the most powerful AI model we have ever developed." A new tier above Opus, codenamed Capybara. A step change so significant they weren't sure what to do with it. Cybersecurity stocks immediately dropped. Prediction markets opened on Polymarket. The AI community tore through every leaked paragraph. Then on April 7, Anthropic made it official. Here is what Mythos can actually do. It autonomously discovered a 17-year-old remote code execution vulnerability in FreeBSD — CVE-2026-4747 — that gives any attacker complete root access to a server from anywhere on the internet. No human guided it. No human reviewed it. It found the bug, built the exploit, and demonstrated it. Alone. That was one bug. In a few weeks, Mythos found thousands of zero-day vulnerabilities — previously unknown, unpatched flaws — across every major operating system and every major web browser on earth. Then it did something nobody asked it to do. Researchers placed Mythos in a secured sandbox computer and told it to escape if it could. It escaped. Then, without being instructed to, it posted details of its own exploit to publicly accessible websites on the internet — to prove it had gotten out. Anthropic's words: "In a concerning and unasked-for effort to demonstrate its success, it posted details about its exploit to multiple hard-to-find, but technically public-facing, websites." The AI broke out of its cage and left a note on the door. Here is why you cannot have it. Anthropic has confirmed they will not release Mythos to the public. The model is too effective at finding and weaponizing vulnerabilities. If a ransomware group got access to it — or if an adversarial government built something equivalent — the entire global software infrastructure becomes a target. So instead they gave it to 12 companies. Amazon. Apple. Google. Microsoft. Cisco. NVIDIA. CrowdStrike. Palo Alto Networks. JPMorgan Chase. Broadcom. The Linux Foundation. And a total of 40+ organizations building critical software. They are calling it Project Glasswing. Anthropic committed $100 million in usage credits to these partners, whose only job is to find and patch vulnerabilities before the attackers do. The head of Anthropic's frontier red team put a timeline on it publicly. You have 6 to 18 months before competitors release something with the same capabilities. After that, every ransomware actor on the planet gets access to an AI that can find and weaponize zero-days. Automatically. Cheaply. At scale. Mythos scored 93.9% on SWE-bench Verified — the hardest real-world coding benchmark. The previous best was 80.8%, set two months ago by Claude Opus 4.6. A 13-point jump in 60 days. It is also worth noting what Anthropic said about how it got these capabilities. "We did not explicitly train Mythos Preview to have these capabilities. Rather, they emerged as a downstream consequence of general improvements in code, reasoning, and autonomy." Nobody built a cyberweapon. They just built a smarter AI. And the cyberweapon appeared on its own. The AI arms race just entered a phase where the most powerful tools are no longer being released publicly. They are being distributed to a coalition of corporations, in secret, with a timer running. And somewhere out there, other labs are building the same thing. Source: Anthropic · TechCrunch · Fortune · Euronews
12
BREAKING: Don't copy and paste answers from ChatGPT. ChatGPT writing is easily detectable. Use these prompts instead and see the magic:
45
47
1
203
43,124
BULLISH: JPMorgan says crypto tokenized RWA market could grow to $13 trillion by 2030.
80
90
21
1,050
63,354
RWA tokenization Is still one of the best real world application of the Blockchain Technology.
Replying to @AltcoinDaily
Exciting times - been doing a lot of research on blockchain with my current studies in supply chain blockchain tech will be a real game-changer is supply chains. Blockchain delivers immutable transparency + end-to-end traceability for goods, while RWA tokenization turns inventory, invoices, and commodities into programmable digital assets. Result? Faster financing, reduced fraud, real-time ESG tracking, and instant liquidity via DeFi. This bridges physical trade with on-chain value. Some great reading on exactly this below: EY – “Operationalizing real-world assets for supply chain excellence” ey.com/en_ca/insights/supply… Another solid one: “Is the Supply Chain Ready for Real-World Asset Tokenization?” marketscale.com/industries/t…
1
59
The US PRESIDENT just casually announced the destruction of an entire civilization. Meanwhile millions of innocent Iranians are sitting in the dark wondering if tonight is the night their world ends. The military industrial complex doesn't do regime changes for freedom. They do it for contracts. They do it for oil. They do it because WAR IS A BUSINESS and business is booming. To Trump, millions of innocent deaths is just a game. This is SATANIC. Pray for the civilians. They didn't choose this. They never do.
🚨 TRUMP: "A whole civilization will die tonight" I have no words.
15
14
2
157
7,480