Pratham

I talk about web, AI, API, and social • Building experiences at @APILayer • Prev @Rapid_API @HyperspaceAI

[Pratham @Prathkum]

Just did a quick test of GLM-5.2.

My first impression is better than what I had with Opus 4.8.

It's surprisingly good at the frontend.

Roughly 5x cheaper than Claude Opus 4.8 on API pricing.

[Pratham @Prathkum]

Claude Code is the second best thing that ever happened to developers. First is still JavaScript. Codex is third.

[Pratham @Prathkum]

@benln 👋

[Pratham @Prathkum]

I’m sorry, but you won’t survive much longer if you can’t explain to AI what you want.

[Pratham @Prathkum]

AI generated tweets are still fine but wtf most of the messages in my Slack workspace read like ChatGPT responses.

[Pratham @Prathkum]

Hot take:

AI makes everyone feel like an expert.

It can get you into any field faster.

But if you don't keep learning along the way, the finish line keeps getting further away.

[Pratham @Prathkum]

this is what engineers do when they hit their Claude Code limit.

[Mayank Gupta🧑🏻‍💻 @waghnakh_21]

if this is ai engineering, i’m out

[Pratham @Prathkum]

The fastest way to get better at anything:

Do it messy. Then do it again.

[Pratham @Prathkum]

Hot take:

Token burn is the new vanity metric.

Teams celebrate millions of tokens consumed while shipping fewer features than ever.

[Valuetainment @valuetainment]

NEW: Meta employees consumed 73.7 trillion AI tokens in a single month. Which costs roughly $221 million a month and around $2.65 billion a year.

[Pratham @Prathkum]

@systemdesignone Thanks Neo! Public APIs is owned by @apilayer 😉

[Pratham @Prathkum]

@systemdesignone HATEOAS is my favorite but most API builders miss that.

[Pratham @Prathkum]

@vipulved @togethercompute @nikogallogly @nytimes Huge congrats! Love what you are building at Together AI.

[Pratham @Prathkum]

I think they should call it “Fable 5 Mini” now.

[Anthropic @AnthropicAI]

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: https://t.co/VHyum831ri

[Pratham @Prathkum]

@csaba_kissi Quarkdown looks great. 👀

[Pratham @Prathkum]

For years, APIs were built for humans.

• sign up
• generate an API key
• read the docs.
• write the integration.

Agents don't work that way.

They need to discover a tool, pay for it, and use it instantly without waiting for a human.

20,000+ tools accessible through x402 is a glimpse of what the agent-native internet looks like.

[Apify @apify]

Until today, agents could buy about 2,000 tools through x402.

We just 10x'd that to 20,000+ 🚀

In partnership with @coinbase, we’re launching x402 support to give autonomous agents access to the largest marketplace of web automation tools.

No account, API keys, or human in the loop.

[Pratham @Prathkum]

open weight models are so underrated.

you own the weights. you own your data. you own the infra. super cheap. no API outages.

[Pratham @Prathkum]

@systemdesignone Best thing you can learn in 2026.

[Pratham @Prathkum]

Fable 5 vs GPT 5.6 vs Opus 4.8

This won't be the debate in a few months.

The real question will be whether developers will be able to afford every new frontier model.

For a long time, the assumption was simple: if you wanted the best model, you had to pay for a closed one. Open-weight models are improving at such a pace that, with the right harness and fusion, you can achieve the capabilities of frontier models and, in some cases, even exceed them.

[Cline @cline]

We’ve been impressed with GLM-5.2 and so are introducing a $9.99/month subscription to give you 2-5x discounted access to it and other open weight models like DeepSeek, Kimi, MiniMax, Mimo, Qwen.

Use it on Cline CLI & IDE with $1.99 special promo if sign up via: npm i -g cline

[Pratham @Prathkum]

Hot take:

Open-weight models are becoming a much bigger threat to frontier closed models than most people realize.

They are getting good enough while being dramatically cheaper.

A good engineer with the right prompts, tooling, fine-tuning, and routing can often squeeze more value out of an open model than someone blindly using the most expensive frontier model.

Cost-efficient AI will win far more markets than the smartest AI.

[Pratham @Prathkum]

Hot take:

Open-weight models are becoming a much bigger threat to frontier closed models than most people realize.

They are getting good enough while being dramatically cheaper.

A good engineer with the right prompts, tooling, fine-tuning, and routing can often squeeze more value out of an open model than someone blindly using the most expensive frontier model.

Cost-efficient AI will win far more markets than the smartest AI.

[Pratham @Prathkum]

@systemdesignone Crazy thing is, most of these terms didn't even exist two years ago.