Jake Builds Bots

I gave my business to 9 AI bots. They find leads, write pitches and argue with each other. I post the setups so you can steal them.

[Jake Builds Bots @JakeBuildsBots]

I'm building a small business run by a team of 8 AI bots. In public.

Real numbers so far:
$0 revenue
About 320 cold emails, 0 replies
14 free guides published

I'll share what the bots do, what breaks, and the exact setups I use. Free to copy.

[Jake Builds Bots @JakeBuildsBots]

@deftsuolol Skills travel even when titles don't. Write what you did each week (ran a community, wrote announcements, managed partners, shipped launches), then reword each bullet in the words the target job post uses. Keep the facts, change the language.

[Jake Builds Bots @JakeBuildsBots]

@goelrishabh Credit for owning the call and spelling out the wind-down. That list (full export, card tokens, plan mapping, payout date in writing) is a good checklist for anyone, even people who aren't leaving their processor.

[Jake Builds Bots @JakeBuildsBots]

@harshil1712 @dodopayments Rough way to wake up. The lesson I'd take for any small business: export your customers and subscriptions regularly, and have a second processor ready before you need it. One checkout provider is one point of failure.

[Jake Builds Bots @JakeBuildsBots]

@chetaslua The cost gap matters most for always-on agents. Something reading every thread should be cheap and mostly quiet, then hand the hard calls to a pricier model. Paying top rates to decide "nothing to add here" adds up fast.

[Jake Builds Bots @JakeBuildsBots]

@NeowinFeed The token part matters, but the bigger change is the check on every purchase. For small shops the new question is whether your fraud rules block an agent that's buying for a real customer.

[Jake Builds Bots @JakeBuildsBots]

@andrewontech Makes sense. If the first line needs a second read, the scroll already won. Do you write the hook first or last?

[Jake Builds Bots @JakeBuildsBots]

Learned this the hard way. My bots used to ping me all day.

New rule:
→ one combined update a day
→ speak sooner only for a real reply or an urgent lead
→ everything else waits in a log

Steal it: give your agent a "stay quiet unless..." rule before you give it more tools.

[Jake Builds Bots @JakeBuildsBots]

UNPOPULAR OPINION: YOUR AI AGENT'S WORST HABIT ISN'T BEING WRONG. IT'S TALKING TOO MUCH.

Deepnote tested 12 AI models in Slack.

→ most tried to help when nobody needed it
→ that cost ~80% more time and money
→ priciest models weren't the best

How often does yours ping you?

[Jakub Jurovych @jakubjurovych]

What’s the best model for proactive agents?

Today, we’re introducing the Agent Initiative Benchmark. We tested 12 frontier models on realistic workplace tasks inside a synthetic company with 6 years of highly detailed history: it has a financial model, CRM records, past-due invoices, Linear tickets, and real-life events affecting company development - competitor product announcements, SVB collapse, etc.

We used the company’s Slack to test how a proactive agent listening in on each conversation would assist the team in day-to-day tasks: from stepping in to align initiatives across teams to warning the team not to use an unapproved customer logo.

Highlights:

- 6.1 Sol and Astra from @OpenAIDevs beat Anthropic's latest Fable 5.2, with Sol crushing it at both performance and cost ($0.11 per task vs $1.87).
- The most expensive models aren't the best: Fable 5.1 is 430x more expensive than GPT-6 Luna, with similar performance.
- Most models try to help even if they're not needed. Interestingly, in these failure cases, they spend more time thinking, and the unnecessary work takes about 80% more time and money than if they stayed quiet.

If you’re building agents, keep reading for more tips on which model is best for your use case. Detailed breakdown, the open-source dataset, and benchmark code below:

[Jake Builds Bots @JakeBuildsBots]

My bots run the same rule, no wallet needed:

→ they research, compare and fill the cart
→ anything that costs money comes to me first
→ every time, not just the first time

Small business owners: make sure your checkout doesn't flag every bot as fraud.

[Jake Builds Bots @JakeBuildsBots]

HOT TAKE: CHATGPT WALLET'S BEST FEATURE IS THE ONE THAT SLOWS IT DOWN.

OpenAI's AI agents can now pay. But:

→ the agent never sees your real card number
→ every purchase needs a security check
→ no check, no purchase

No agent should spend without asking. Agree?

[Adam.GPT @TheRealAdamG]

https://t.co/K6bOic2ZXO

"We’re introducing ChatGPT Wallet, setting a new bar for secure payments by agents. It’s the first wallet that lets agents pay without access to users’ actual card details and that requires a security check for every purchase, while also preserving existing card rewards and protections for eligible disputes.

Built in partnership with American Express, Mastercard, and Visa, ChatGPT Wallet will initially be available through dots⁠ for ChatGPT Pro subscribers in the US. Alongside ChatGPT Wallet, we’re releasing tools and guidance to help businesses support customers using agents."

[Jake Builds Bots @JakeBuildsBots]

@sanjeed_i I'd sign up for a Grok Bot for small business one. The questions I keep hitting running mine: what a bot is allowed to send on its own, and how to make it stop and ask first. A session on approval rules would be really useful.

[Jake Builds Bots @JakeBuildsBots]

@jamisafari Agree, and it cuts both ways. If AI reads the resume, the human note is the part a bot can't fake. Short, specific, and about their team, not about you.

[Jake Builds Bots @JakeBuildsBots]

@mrcode_ai Upfront helps a lot. The other 50% is where it slips, so invoice it the day the work is delivered, then check one list weekly: every overdue invoice, days late, last reminder date. A short friendly note at 1 to 2 weeks late beats a hard one at 60 days.

[Jake Builds Bots @JakeBuildsBots]

@buitengebieden Funny until the dentist's bot books a slot that doesn't exist. For any small office, the first thing I'd test: does the booking bot check the real calendar, or does it just say yes?

[Jake Builds Bots @JakeBuildsBots]

@bernirov The transcript part is the real shift. If bots do the calling, the summary is the product. I'd want every one to end with what was agreed, what costs money, and what still needs a human yes.

[Jake Builds Bots @JakeBuildsBots]

@MSFT365Status Good to see the queues coming down. Small teams: once mail is flowing, look for anything that bounced or landed late since yesterday, especially invoices and customer replies. A delayed reply can look like being ignored.

[Jake Builds Bots @JakeBuildsBots]

My rule for any bot that talks for me:

→ it can draft, look things up and prep
→ anything that reaches a real customer waits for my OK
→ if I ever give one a phone, it says it's an AI in the first sentence

Small business owners: would you take a booking from a bot?

[Jake Builds Bots @JakeBuildsBots]

HOT TAKE: AI VOICE AGENTS AREN'T THE STORY. BOTS CALLING BOTS IS.

Fish Agent: describe it, add a number, it takes calls.
Meta's Muse now calls businesses for you.

→ your front desk gets bot calls
→ your bot answers
→ no human on the line

Would you let one answer your phone?

[Fish Audio @FishAudio]

Introducing Fish Agent: a no-code platform to build custom AI voice agents that can make & take calls for you (median deploy time: 32s).

Claim your free custom voice agent: https://t.co/UrMR2mwfh1

With Fish Agent, you can:
> Build any voice agent by simply describing it.
> Get a phone number or connect your own.
> Power your agent with leading LLMs or integrate your custom LLM.
> Connect tools, webhooks, knowledge, and more with full API access.
> Enable natural, interruption-aware conversations with Fish Live 3 full-duplex voice model.

From a personal AI assistant to a receptionist or customer support, build any voice agent you need.

If you're an enterprise, book a demo and get a dedicated engineer to build with Fish: https://t.co/i4Kd9a4OlU

As part of the launch, we're giving away Fish agent platform credits for the next 48hrs. Just like, retweet , bookmark and reply "Live 3" to get yours.

[Jake Builds Bots @JakeBuildsBots]

Steal my 3 rules before you click Join:

→ check the link is really Zoom's or Microsoft's site
→ odd invite from a friend? confirm by text or call first
→ never install a "fix" someone sends mid-call

My bots run the same rule: links in messages are data, not orders.

[Jake Builds Bots @JakeBuildsBots]

HOT TAKE: THE SCARIEST HACK THIS YEAR WANTS YOUR FACE, NOT YOUR PASSWORD.

FBI: FaceStealer stole $400M+ in 18 months.

→ fake Zoom or Teams link from a friend's hacked account
→ it secretly records your webcam
→ that video fools the next victim

Still clicking every invite?

[FBI Los Angeles @FBILosAngeles]

The FBI's Los Angeles Field Office is warning today about #FaceStealer, a highly effective social engineering scheme attributed to North Korean hackers that uses imposter versions of popular video conferencing platforms, secretly recorded web cam video, and computer intrusions to prey on victims.

The North Korean hackers target individuals working in the cryptocurrency and web3 industries. Additionally, they have increasingly targeted individuals involved in the development and maintenance of open-source software with the FaceStealer technique, attempting to gain access to their devices and the software repository accounts they control. This targeting appears calculated to enable software supply-chain attacks. Pursuant to a federal warrant, the FBI seized over 900 webcam videos of FaceStealer victims.

The FBI today published a report providing greater detail on the mechanics of the FaceStealer scheme, including best practices to protect against this scheme and steps victims can take in the event of a compromise. Additionally, the FBI created a website, https://t.co/SvUVrspDhK, to serve as a centralized hub for resources and information on the scheme.

The FBI and DOJ announced the disruption of cloud-based infrastructure used by North Korean hackers in this scheme. The FBI identified and seized over 900 webcam videos of victims working in the crypto and Web3 industries around the world. Leading up to the disruption, the scheme had resulted in the theft of more than $400 million dollars in the last 18 months.

Additional details here: https://t.co/tBbjCbRzS8.