Ethan Lo

49 posts

Ethan Lo

@ethanhlo

ai @adaptivesec | co-founder @attentivemobilehq | ex-twitter | civic enthusiast

NYC Katılım Mart 2020

113 Takip Edilen30 Takipçiler

Ethan Lo@ethanhlo·1d

The reason AI-powered fraud converts better isn't sophistication for its own sake. It's that personalization at scale was previously impossible. An attacker could spend 4 hours crafting one perfect spear phish, or use AI to send 10,000 of them. The economics changed. The training has to change too.

Adaptive Security@AdaptiveSec

According to a recent @INTERPOL_HQ report, AI-powered fraud schemes are estimated to be 4.5x more profitable for criminals than traditional tactics. Legacy security awareness training wasn't built for this threat landscape. @AdaptiveSec is.

English

Ethan Lo@ethanhlo·1d

@levie we built Salesforce read/write into our internal OS and immediately started finding new use cases

English

190

Aaron Levie@levie·2d

Agents are going to use software 100X more than people will in the future. As a result, enterprise platforms will become headless and be able to work with any agent on or off platform. If you don’t do that you’re DOA. What some have missed is that this creates vastly more use-cases for these platforms than even existed pre-AI. This isn’t zero sum. Software value props have traditionally been capped at the number of users you have in a company. Agents have no upper limit. We’re going to run agents to process data at a scale humans never could, they’re going to be running 24/7 in parallel doing work for us, and they can integrate workflows across systems to generate all new value propositions. Once you embrace this approach, it becomes obvious how much more upside there is.

Marc Benioff@Benioff

Welcome Salesforce Headless 360: No Browser Required! Our API is the UI. Entire Salesforce & Agentforce & Slack platforms are now exposed as APIs, MCP, & CLI. All AI agents can access data, workflows, and tasks directly in Slack, Voice, or anywhere else with Salesforce Headless 360. Faster builds, agentic everything. 🚀 #Salesforce #Agentforce #AI venturebeat.com/ai/salesforce-…

English

136

177

1.8K

338.8K

Ethan Lo@ethanhlo·1d

@vasuman fully agree if a better model ships tomorrow, we're able swap it in but only because we built our context and tool layer to be independent first

English

413

vas@vasuman·2d

Betting your entire AI stack on one vendor (OpenAI, Anthropic, Google) is the worst decision an enterprise can make right now. Price hikes, quantization, rate limits, migration costs are all coming. You need to abstract your model layer or you'll start to feel the pain soon.

English

364

87.3K

Ethan Lo@ethanhlo·1d

@dwr the compounding part is what people miss. you don't feel the debt until you try to give an agent company context and realize half your institutional memory lives in DMs that don't exist anymore

English

Dan Romero@dwr·2d

Bias for communication in public Slack channels (with thread hygiene!) is a huge, compounding advantage.

TBPN@tbpn

Zapier CEO @wadefoster says remote companies have an advantage in the AI era, becuase "every last bit of work exhaust is documented," which supercharges internal AI and accelerates people's work: "All our stuff is inside Slack. All our meetings are recorded. Every last inch of work that happens, there is a written trace of that." "So we can put chatbots on top of that, and that creates a whole bunch of institutional knowledge that accelerates the work." "So a new person coming in can literally figure out, 'Is there a standard operating procedure for this?' And you don't have to go chase people down in offices and sort of hope the campfire wisdom finds you." "Remote companies have a big advantage because they do tend to have so much work that leaves a digital exhaust."

English

15.2K

Ethan Lo@ethanhlo·1d

@andrewchen challenge being, how do you develop and maintain taste without that middle 80%?

English

andrew chen@andrewchen·2d

observation- for workflows using AI, you spend LESS time for the middle 80% of the work -- instead, you spend MORE of the saved time on the first 10% and last 10% where taste expresses itself that is, generating ideas to start, and validating at the end. And iteration

English

137

8.8K

Ethan Lo@ethanhlo·2d

@ConnerBean @fatih super-superpowers is building your own mcp server and giving it access to anything you can dream of :)

English

Conner Bean@ConnerBean·3d

@fatih Add Linear on top, and the velocity is insane!

English

453

Fatih Arslan@fatih·3d

Superpowers is giving your agent MCP access to Slack and Notion.

English

13.7K

Ethan Lo@ethanhlo·2d

@plainionist Bad specs used to produce bad software slowly, now they just produce bad software fast

English

Seb@plainionist·3d

Hot take: Software engineers will soon write more specs than they read. Specs for acceptance criteria. Specs for architecture. Specs for everything. AI handles the implementation.

English

5.8K

Ethan Lo@ethanhlo·2d

@claudeai wow - and wow that mythos is on here, y'all can't help teasing us huh

English

2.7K

Claude@claudeai·2d

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.

English

4.7K

10.3K

80.9K

12.9M

Ethan Lo@ethanhlo·3d

@remondimi @OpenAIDevs @temporalio being able to switch between models any time there's an incident is absolutely table stakes

English

Mike Remondi@remondimi·3d

We built our own Agent Builder about a year ago on top of @temporalio and deeply integrating it with our own tool calls and data. This is interesting... but the ability to switch to any model provider and build our own functionality on top of it still seems too valuable to switch to something like this

English

536

OpenAI Developers@OpenAIDevs·3d

Build long-running agents with more control over agent execution. New capabilities in the Agents SDK: • Run agents in controlled sandboxes • Inspect and customize the open-source harness • Control when memories are created and where they’re stored

English

242

2.3K

374.8K

Ethan Lo@ethanhlo·3d

@AdaptiveSec The same voice cloning tools developers use to build cool demos are the ones being used to scam elderly people out of their savings. The technology is neutral. We have to move faster than the people using it for this.

English

Adaptive Security@AdaptiveSec·3d

Older Americans lost $81 billion to fraud in 2024. We sat down with Brady Finta, a former FBI special agent and founder of the National Elder Fraud Coordination Center, to learn more about this issue.

English

150

Ethan Lo@ethanhlo·3d

@zeeg i mean this is proof of product market fit tho

English

David Cramer@zeeg·5d

anthropic being down, rate limits, etc is understandable. huge hardware reqs and insane growth. claude code being bug filled is just a result of people vibe coding slop at anthropic and accepting it as ok. it is not ok you will lose your customers just as fast as you've gained them.

English

1.1K

117.6K

Ethan Lo@ethanhlo·3d

@Polymarket workslop is very real: biggest issue we see is people sending ai-generated stuff to colleagues without reviewing it first. the burden has to be on the sender

English

893

Polymarket@Polymarket·4d

JUST IN: Use of AI in the office is reportedly creating a flood of “workslop” that takes longer to fix than do from scratch.

English

351

701

9.4K

1.2M

Ethan Lo@ethanhlo·3d

@staysaasy the best attitude is what @tryramp suggests and what we're adopting - don't get attached to tools, get attached to fixing problems

English

staysaasy@staysaasy·4d

It’s impossible these days to sound like you’re ahead of the curve on AI. Every time I talk to someone about what their company does, it sounds like they’re behind schedule. The only way to sound ahead of the curve is to have completely changed your company to adapt to a thing that got released like last week. I find this concerning.

English

6.9K

Ethan Lo@ethanhlo·4d

great read - "We don't always need to understand the machine to drive the car." the end state: an interface as simple as a steering wheel, shifter, and pedals.. where anyone can get into any car and drive

Jeff Huang@jeffzxh

x.com/i/article/2039…

English

Ethan Lo@ethanhlo·4d

Mission-critical skills should not be in Claude Desktop — they should be built into production infrastructure. Learned this the hard way. The gap between "this Claude workflow is impressive" and "this Claude workflow is reliable" is enormous, and almost nobody is talking about it.

English

Ethan Lo@ethanhlo·9 Nis

we've been building real workflows on top of Cowork at @AdaptiveSec — support automations, program reviews, renewal prep — and it's been surprising how far you can push it with skills + MCPs. let's goooo

Claude@claudeai

Claude Cowork is now generally available to all paid plans. For Enterprise, we are adding role-based access controls, group spend limits, usage analytics, and expanded OpenTelemetry to give admins what they need to deploy it across the org.

English

Ethan Lo retweetledi

Mike Remondi@remondimi·4 Nis

x.com/i/article/2040…

ZXX

401

Ethan Lo@ethanhlo·6 Nis

Next thing we're prototyping at @AdaptiveSec: a semantic layer so LLMs actually understand our internal vocabulary. "Campaign" means something very specific in our product — generic LLMs don't know that, and it causes real quality issues in outputs. The fix is weirdly simple and I think most AI-enabled companies are going to need one.

English

Ethan Lo@ethanhlo·1 Nis

Gong and Salesforce both shipped to prod at @AdaptiveSec this week. Two days after Gong went live, the team had already run 579 Gong searches through Claude. Zero to real usage that fast. So proud of the work

English

Ethan Lo@ethanhlo·26 Mar

Look, I love @claudeai and their pace of development is awesome (what like 10 products in the last two weeks??) but also their pace of bugs/feature gaps is also "awesome": - basically daily morning incidents - Claude Code on web/desktop has no Open Telemetry support - Direct invocation of skills in Cowork is not included in OTel - Github MCP is broken on Cowork - Organizational Plugin Updates were broken for weeks Let's hope they're fixing as well as building..

Claude@claudeai

Your work tools in Claude are now available on mobile. Explore Figma designs, create Canva slides, check Amplitude dashboards, all from your phone. Give it a try: claude.com/download

English

155

Keşfet

@levie @vasuman @dwr @andrewchen @ConnerBean @fatih @plainionist @claudeai