Co-founder & CEO at @goairspeed. Companies as code: agents, evals, memory, and recursive self-improvement. Ex-DeepMind.
London, England
Joined January 2010
- Tweets1.7K
- Following610
- Followers467
- Likes2.9K
An early @goairspeed advisor told me to treat every founder hour as if it cost $1,000.
As an engineer, I think of it simply: founder time is a scarce resource. Spend it where your involvement actually changes the outcome.
AI changes the economics: a small harness around Codex (sorry, ChatGPT) can now take on surprisingly context-heavy "$20 tasks."
I wrote about the system I built on top of @DiogoSnows's open-source Glide and what I've learned: adliska.com/2026/08/25/your-…
This shouldn't be limited to founders and builders. Reps, managers, CROs, and everyone else on a GTM team should have a personal AI assistant that understands their context and takes work off their plate.
We're bringing this experience to Airspeed soon. Stay tuned.
Your forecast spreadsheet says what the team believes. Buyer evidence tells you whether it holds up.
Last week we launched file attachments in every @goairspeed agent chat.
Drop in the workbook and ask:
- Which committed deals have weak buyer evidence?
- Where does the rep’s confidence disagree with recent calls?
- Which three deals need manager attention this week, and why?
The agent assesses the spreadsheet alongside the CRM data, buyer conversations, and account history Airspeed already has.
No copying rows into prompts. No asking a generic assistant to judge a deal without the context.
The same applies to the other messy artifacts GTM teams actually run on: proposal decks, pricing PDFs, dashboard screenshots, and account CSVs.
A few other things we shipped last week:
- Recording-consent pages with customer branding and multiple languages
- Full email access inside Global Ask Airspeed, for both search and drafting/sending emails.
- Improved agent history and chat, including filters for deal-stage-triggered runs
- The full agent toolkit inside Global Ask Airspeed, so the same capabilities are available everywhere
And yes, you probably shouldn't be running your forecast in a Google Sheet in the first place.
That's what Airspeed Forecast is for :-)
But until then, the spreadsheet can at least get a properly informed second opinion.
I love simple initiatives like the London Startup Map. @goairspeed is now on there, right in Holborn:
londonstartupmap.com/company…
@devang314 and I leaned on our network a lot in the early days of Airspeed. Founders, operators, investors, and friends all helped us move faster, pressure-test decisions, and avoid a lot of mistakes.
Anything that makes those connections easier is a big win for the local ecosystem, especially with the momentum in London right now.
If you're building here, add yourself. And thanks for doing this, @fooming86!
Hampstead Heath, London, right now. Thousands gathered to witness a once-in-a-generation event -- the last person in tech changing their title from "sales" to "GTM."
It took me a while to get into @AcquiredFM, but @devang314 talks about it so much that I eventually caved in.
Two ideas from the @Vanguard_Group episode felt quite relevant to me.
(1) Bogle's constraints - low fees, simple funds - built Vanguard. Later he fought new products, especially ETFs. His successors kept the mission - low-cost investing for customers - but stopped treating Bogle's original product choices as sacred. And 99% of Vanguard's AUM came after he stepped down.
(2) You can buy Vanguard ETFs through Fidelity. Vanguard gets distribution; Fidelity owns the interface and customer. Feels a lot like MCPs with Claude and ChatGPT. But Fidelity still offers Vanguard because customers ask for it. The interface has leverage, but so does a product users insist on.
Anyways, a really good episode!
.@MacfarlaneJamie and @CreatorFund were among the first investors in @goairspeed.
Jamie believed early in @devang314 and me, and in our vision for how companies should be run: much more AI-first, data-driven by default, with AI creating huge execution leverage across the business.
We're still some way from that ultimate vision. But Airspeed is already closing one of the most important loops: turning sales conversations, a treasure trove of customer intelligence, into action across the whole company.
That means improving how teams sell, sharpening how they message the product, and feeding what customers are saying directly into the product roadmap.
A few years later, we found ourselves crossing London in an Airspeed-branded taxi, talking about the journey so far. (Who would've thought we'd do this back in the day! 😅)
Jamie, thank you for believing early. We're lucky to have had you in our corner from the start.
I got a cold email this morning from @mercor about licensing @goairspeed's codebase to train frontier models. Potential compensation: up to $200k+.
The interesting part is why our code is valuable.
Airspeed's codebase encodes how GTM teams operate: how deals progress, where they stall, which signals matter, what action should happen next, and how all of this maps across CRM, calls, email, and Slack.
Public code is abundant. Executable models of how real business functions operate are not.
This is becoming an explicit market. Mercor licenses enterprise workflow data, while @turingcom offers up to $1M for a company's operating history.
Process knowledge used to live in employees and playbooks. Increasingly, it lives in software. A playbook describes how a company works. A codebase executes it.
That code isn't exhaust to license once. It's proprietary infrastructure that can improve how your company operates every quarter.
Keep it. Govern it. Compound it.
You gotta appreciate a good cold email.
This one from Simon at @rhobusiness caught my eye. Simple. Four bullets. Proof. A phone number. One question + link.
So I tried it for @goairspeed :
Subject: why revenue teams move to airspeed
Hi Jennifer,
Gave you a ring. Revenue teams move from @Gong_io and @clarihq for four reasons:
- Catch: deal risk flagged weeks early
- Fix: every deal babysat (prep, nudges, follow-ups, CRM)
- Forecast: every number traceable to buyer evidence
- Service: RevOps analyst + my direct line on every deal
30% shorter sales cycles (@useascend). Live across 100+ reps in days (@Price_fx). Free until your current contract ends.
Open to chat this week?
Most sales forecasts are vibes wrapped in a spreadsheet.
Forecasting isn't an LLM summarization problem. It's an evidence problem.
We're demoing @goairspeed Forecasting live in 30min: every number traceable to CRM state, buyer engagement, and historical outcomes.
goairspeed.com/webinars/sale…
Adam Liska retweeted
We've been quiet here for a while. That's changing.
Earlier this year, Glyphic became Airspeed.
A few weeks later, we announced our $20M Series A, led by @DNCapital, with @frameworkvc , @VI_Partners_AG and @Atlassian Ventures.
Here's how we got here: goairspeed.com/blog/the-stor…
Adam Liska retweeted
Two ex-DeepMinders have just raised a $20m Series A to build the future of GTM
The company, founded by @adliska and @devang314, has built an agentic GTM platform.
Its AI agents act on context across calls, emails, tickets, and CRM, automatically updating systems, flagging risks, and generating follow-ups to help sales teams close the deal.
The company now has over 200 clients across 20 countries and has 4x'ed revenue in the last 12 months.
It has just announced its $20m Series A led by DN Capital.
AMAZING NEWS - the Google DeepMind mafia is going strong 👏
Adam Liska retweeted
Something tells me even the Grinch will stand up for an encore. #RAHChristmas
Evals will be an interesting next frontier. There's quite a bit happening in the monitoring/analytics space for models in production, but evals are still often developed and kept in-house. They also get more difficult as capabilities of AI models increase.
Always look at your data!
One pattern I noticed is that great AI researchers are willing to manually inspect lots of data. And more than that, they build infrastructure that allows them to manually inspect data quickly. Though not glamorous, manually examining data gives valuable intuitions about the problem.
The canonical example here is Andrej Karpathy doing the ImageNet 2000-way classification task himself. And in the era of large language models, manually examining data is probably even more insightful since completions are hard to evaluate via benchmarks.
In this spirit, I recently did a few days of pair programming with @hwchung27 where we were starting on a new problem. Instead of trying to replicate baselines and design new methods, we ran some evaluations and manually inspected them to gain insights. We first paid about one day of overhead getting all the relevant information in a single UI so we could examine the data without having to click through multiple web pages. The second day, we spent an afternoon reading examples together and taking notes on the patterns that we noticed in the examples. ChatGPT generates long text, and we actually read the whole thing carefully, even if one example took 20 minutes to understand. I think we both gained a deeper understanding of the problem that we could not have gotten from reading research papers.
(In 2018, for example, I helped pathologists label a lot of data to train a lung cancer classifier. After having manually labeled 200+ images (with pathologist correction), I’d probably gained a pathologist-level understanding at that one particular lung cancer classification task :))
Adam Liska retweeted
We can’t wait to see what @GlyphicAI's copilot, coupled with Crunchbase’s best-in-class company data, can help salespeople accomplish! 🚀
Discover what Crunchbase data can do to power your #AI tool: about.crunchbase.com/product…
📢 We are thrilled to announce our strategic partnership with @crunchbase.
@GlyphicAI's sales copilot is getting a boost with access to Crunchbase's top-tier company data, enabling smarter decisions throughout the sales cycle.
Read more: glyphic.ai/post/glyphic-and-…
I'm in New York this week before flying off to Nashville next week for @Join_Pavilion's GTM2023 event.
Let me know if you're in the New York area or attending GTM2023 and want catch up for a quick ☕!
ALT New York mural graffiti
I'm very excited about this partnership with @crunchbase and what it'll bring to our users!
📢 We are thrilled to announce our strategic partnership with @crunchbase.
@GlyphicAI's sales copilot is getting a boost with access to Crunchbase's top-tier company data, enabling smarter decisions throughout the sales cycle.
Read more: glyphic.ai/post/glyphic-and-…
Adam Liska retweeted
Wow, what a success! 💥
We had a full house for our #GenerativeAI meetup, co-organized with @NotionCapital and @nvidia.
Our CTO @devang314, Radu Bozga, and @grzywada talked about opportunities and challenges of building AI-native products - and sparked a lot of conversations!
