Fractional CTO & AI builder. I help startups ship & fix the real technical bottleneck. EU/UK/US. Free AI audit call: https://teapawt.net/t.co/1oAzRf3C8s
Joined January 2025
- Tweets17.9K
- Following2.4K
- Followers1.8K
- Likes8.3K
Pinned Tweet
Most "we need to hire a CTO" moments aren't about hiring.
9 times out of 10 it's an unclear technical priority + too much context-switching, not missing headcount.
Map the ONE constraint blocking shipping, fix that first.
Fractional CTO here, happy to sanity-check your stack.
Yash Vardhan retweeted
Introducing Kimi K3: Open Frontier Intelligence
🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows
Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.com/blog/kimi-k3
Yash Vardhan retweeted
Claude Fable 5 will be available again globally tomorrow.
After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.
We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.
Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.
Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.
Read our full blog: anthropic.com/news/redeployi…
The best technical decision is usually the boring one.
Proven stack. Simple architecture. Visible failure states.
"Boring" ships and survives.
"Clever" gets a standing ovation and a 3am outage.
Yash Vardhan retweeted
BREAKING: SpaceX has agreed to acquire Cursor, the world's fastest growing software startup, for $60 billion in an all stock deal.
Cursor has over 1 million paying customers, more than $2 billion in annualized revenue, and is projected to hit $6 billion by end of 2026.
At $60 billion, this is the largest software acquisition in history, paying 20 to 30 times Cursor's current revenue.
The deal is subject to regulatory approval and expected to close in Q3 2026.
SpaceX now owns the rockets, the satellites, the AI models, the chips, and is about to own the tool every developer on earth uses to write code.
Every founder underestimates the same thing: not how hard the build is, but how much context-switching quietly destroys their team's output.
Protect focus like it's runway. Because it is.
Most startups don't need to hire a CTO. They need to fix one constraint.
When shipping feels slow, the instinct is 'we need senior technical leadership.' Usually that's not the actual problem.
The actual problem is one unclear priority. Nobody’s decided what 'done' means for the next thing, plus a founder context-switching between sales calls and code review.
A full-time CTO is a nine-month, six-figure bet to fix something a five-day diagnostic could name.
Hire the CTO when you have a technical org to lead. Until then, find the constraint and fix that.
It was fun while it lasted! 🥲
If Fable 5 disappears, what’s your replacement for long-running agent workflows?
Haven’t found anything that feels quite the same yet.
Back to the model roulette wheel we go. 😂
The gap between teams is about to widen, not shrink.
Give a model this capable to a team with a clear technical priority → they 10x.
Give it to a team that's still arguing about what to build → they generate 10x more of the wrong thing, faster.
A model that's state-of-the-art on every benchmark still ships nothing without:
- A clear spec
- Visible failure states
- A human gate on the irreversible step
Fable 5 is a better engine. You still have to decide where the car is going.
With every release the distance between an idea and a shipped business gets shorter. The tools are becoming multipliers, clear thinking is becoming the moat.
Replit’s evolution is top notch.🔥 @amasad
Fable 5 is genuinely incredible. It also won't save a startup that hasn't decided what to build.
A better model makes a clear plan faster and a confused plan more expensively wrong. The model got 10x better. The need for judgment didn't move.
Anyone else getting this with Fable 5? 👇
The model feels ridiculously capable, but I’ve had normal agent workflows get flagged and downgraded to Opus 4.8.
Also curious:
How many tokens are people burning on longer swarm-style tasks?
Hiring a full-time CTO to look mature to investors is like buying a Ferrari to look fast in traffic.
Expensive, underused, and solving the wrong problem. Get the right decisions made first. Title later.
I’ll be testing @getcontextdev and this is one of the most practical tools I’ve seen recently for building AI agents that actually interact with the web.
30,000 free credits is enough to:
• Crawl websites at scale
• Build lead-generation workflows
• Enrich company data
• Create AI research agents with live web context
• Automate market intelligence
If you’re building in AI, this is worth claiming.
Reply to @mynameisyahia ‘s post with your use case and he’ll add the credits.
The builders who move fastest are usually the ones experimenting the most. 🚀
30,000 free credits for everyone who replies to this post. 🚀
With 30K credits on @getcontextdev, you can:
> Crawl 30,000 web pages
> Enrich 3,000 companies
> Extract data from 3,000 websites
> Power AI agents with real-time web context
> Build something cool i haven't thought of
No credit card.
Just sign up, reply below with your use case, and I'll add the credits myself.
Let's see what you build 👀
AI agents don't fail because the model is dumb.
They fail at the orchestration layer,
a step firing before its input is ready, no visible failure state, no human gate on the irreversible action.
The model was never the hard part.
#buildinpublic #startups
Technical debt isn't the problem, unmanaged technical debt is.
Every startup takes shortcuts.
The ones that survive write down which shortcuts they took and when they'll pay them back.
The rest just forget, until it breaks in production.