Creator perspective · MoneyRandy Hunt explains that he increasingly calls his team a software design team — a team that understands how software is made and built and designs for software — rather than calling design engineers just designers.
confident
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Report · MoneyRandy Hunt describes Notion's recent investment in a UX writer (a team of one until roughly a year and a half to two years ago), whose first job was to build a UX writing agent to scale themselves and best practices — which also helped people understand what the writer does and when to reach out.
confident
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Reported position · MoneyAlex Lieberman describes Salesforce, as a system-of-record company, embracing a path of providing APIs and MCPs for others to plug into rather than pushing its own agents.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Anecdote · MoneyThe pivot decision was enabled by a hard deadline: roughly a month after October 15th was the only window where they could live in both universes (continuing tax vs. not), and missing it meant waiting a year and losing the capital to change.
anecdotal
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Anecdote · MoneyA business with ~$1M first-year revenue and a projected $2.5-3M second year was the concrete trajectory that, post-AI, the founders judged not venture-scale — a benchmark for what 'good but not venture-scale' looks like.
anecdotal
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Anecdote · MoneyTax was a good market for AI automation because it contains a lot of text, which LLMs handle well, and involves structured input formats like financials and lease documents. The team did technical due diligence and concluded that within two years, AI could automate the process of a small to medium-sized business going from financials and business documents to a finished tax return. The startup generated slightly below one million dollars in revenue in its first year and projected two and a half to three million in year two. Post-AI, that revenue trajectory would not have been venture-scale, whereas pre-AI the same numbers would have been considered a sensible business. Tax software can only be sold once a year, which makes it fundamentally impossible to be a daily-active-use tool and hard,
revenue figures flagged as possible numeric errors
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyCJ Hess explains that all agent logs and documentation at Tenex are checked into Git, marked down and now more HTML — picture books that the agents spit out. Having all context in the repo gives agents and tools superpowers, since a lot of engineering teams still keep that context only in conversations or in their heads. Tenex's goal is to push all context into one place so more and more of the work is driven by the agent rather than provided up front. He also notes there is a debate on whether you even need docs — whether the code can be the memory of a codebase.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Recommendation · MoneyUse AI to prepare for meetings by tweaking decks to match design language and updating them with the latest numbers from PDFs sent to you.
endorsed
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyCarmen Li reports that Compute Exchange has over 100 verified providers and over 80,000 GPUs in supply, covers all regions except China with a 24-hour commitment to round, and sees over a billion RFQs every month ranging from one node to 20,000 nodes. The broader GPU ecosystem now has over 300 cloud providers globally, plus many token factories and API routers.
Figures are the speaker's own claims; some carry possible numeric error flags.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Report · MoneyAlex Lieberman explains that Tenex measures whether its processes are working by tracking the trend in average story points per month of every engineer in the company. CJ Hess reports that overall there has been roughly a 2x increase in engineer output this year, and that over the period from the Opus 4.5 moment to Opus 5.5 he has seen the actual output of engineers roughly double in what they can cover in the same time span. He frames December as the Opus 4.5 moment where models really started to take on long-running tasks. The 2x has also played a similar role in how Tenex does estimations.
approximate figures ('roughly 2x', 'roughly double')
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyMost companies spend roughly $10-200 per employee on tokens; the more AI-native ones convert that to roughly 10%. The average AI startup employee is about 1x in employee spend vs token spend, going up to maybe 5-10x in token spend.
rough estimates from the host
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Principle · MoneyAlex Lieberman argues that agent-as-product companies in specific verticals will struggle unless they have a very strategic moat like compliance or data. Vertical agent products will lose to models as spinning up your own agent gets easier and models continue to improve. However, products that change the process and have an agent on top can still be very helpful and successful. People do not want to use the agent you built — they want to use the agent they already have, plugged into their data.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyAI is really just a leveraged bet on a human's intellect and work ethic — expect more variance in earning capacity than ever before. Go long on yourself: pair a strong software engineering foundation with constant pushing to leverage the frontier and stretch your time while maintaining quality output. There's no better way to earn.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyTenex is somewhere between 1:1 and 2:1 in raw tokens to employee spend. Costs have been driven way down by Opus 5.5; there was an era where everyone used Astra and Fable, which gave the best results but were expensive. A lot of economically valuable tasks can now be done with Opus 5.5.
speaker didn't have exact numbers off the top of his head; possible numeric error in transcript
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyAI is moving so fast that it will change distribution and human habits at a rate where no one can predict the winning platforms, and there will be huge losers. Amazon might lose a huge chunk of its advertising business by trying to fight the future of AI agents, because it cannot predict what that future looks like.
speaker's prediction with moderate confidence (0.6-0.8); explicitly uncertain about outcomes
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyAgent traffic has vastly surpassed human traffic. Amazon has adopted a somewhat adversarial approach to agents, which makes sense because they're an advertising company.
speaker's claims with moderate confidence (0.7)
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyWith a neutral agent, the creator's incentive is to give you the best information for you as a human with no distortion from the great forces of marketing. Users want their AI agent to remain neutral with respect to advertising and marketing forces.
speaker's argument and endorsed claim
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyCompanies that picked OpenAI two years ago then had to use Anthropic too, because Anthropic was better at coding for a 12-month period and they needed those models to be competitive.
speaker's account of market behavior
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Report · MoneyAs the company talks to more restaurant owners, the library of automations grows (from 5 to roughly 15 things for restaurant owners), and AI speeds up the process — scheduling interviews automatically, running AI-recorded interviews with a free Amazon gift card incentive — so they operate at the speed of AI rather than the speed of humans.
described; figures are the speaker's account of their own company
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Principle · MoneyRandy Hunt describes wholly considered experiences as involving logistics and operations but also being very much about feeling — and links whole design with hospitality as very much related approaches.
confident
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small