seb (@sebbsssss) header
seb (@sebbsssss) profile picture

Building Memory @cludeproject Prev ByteDance AI, DBS, StarHub Member @superteamsg

Joined June 2009
Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)Photo by seb (@sebbsssss)
Modern y2k happening. Scary
2
5
477
This is a pretty big deal @Alibaba_Qwen
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery! Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window. Highlights: - Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation. - Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables. - Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery. - $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%. The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder. Start building today! 🚀 - Blog: qwen.ai/blog?id=qwen4-max - API: qwencloud.com/models/qwen4-m… - Token Plan: qwencloud.com/pricing/token-… More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
5
462
Solana 1.0 Nothing else
mainnet-beta → mainnet Introducing Solana 1.0.
1
9
681
Wait till @bot integrates with @XMoney
1
11
885
This jensen saga seems awfully suspicious and coordinated. The first affair is reprehensible but after the second news then the third, that follows after just smells off. which begs the question: who gets to benefit the most here?
1
20
1,375
Matter of time before Grok leads the race. Imo
Replying to @Dimdv99 @tetsuoai
Not as good. Opus 5.5 is a masterpiece. But Grok 5 should exceed Opus 5.5 and maybe that happens in 2 or 3 months, with continued acceleration thereafter.
3
12
1,038
Real men manage their own memory
Real men manage their own memory
3
1
20
1,395
A few people have been asking about my Grok Bot setup. This is one of the skills I actively use. It’s called /optimize. A reusable skill for checking and improving speed, e.g. • N+1 SQL queries → joins or batched reads. • Repeated database reads → caching or Redis where it makes sense. • Serverless connections → connection pooling. • Sequential API calls → parallel requests. • Heavy JavaScript → code splitting and lazy loading. • Blocking tasks → background jobs. • Waiting for a full AI response → token streaming. It measures first, applies safe fixes on a separate branch, then runs the same tests again. You get the before-and-after results and a list of what it couldn’t measure. We used it on a simple internal app we were building. It found a duplicate page load and repeated data reads. Fixing those cut the time from choosing a role to seeing the overview by 32–40% on repeat visits. Free and open source. Setup instructions for Grok Bot, Claude Code and Cursor: github.com/sebbsssss/optimiz…
12
8
3
114
4,788
Disclaimer: This is just a casual sharing posts due to requests, would love to share more in the future
2
360
Take 10 minutes of your time to read. This was written in Oct 23. Fucking eerie, looking at how things are playing out I share the same sentiment that the remainder of 26 and 27 going to be wild...
** IMMINENT STOCK MARKET CRASH OCTOBER 26, 2026-NOVEMBER 9, 2026 ** I am writing this message to give a detailed analysis, as to what I believe will be happening starting Monday October 26, 2026, and the main event to take place the following Monday November 2-3, 2026, with the bottom of the crash coming on November 9. I also want to explain what I believe the Federal Reserve is going to do as a means of trying to fix the crisis, and how it is going to lead to the end of the U.S dollar, bringing about a sovereign debt and currency crisis. This crisis will be the main focal point and issue that the Democratic Party is going to run on in the 26' midterms and the 28' elections, undermining the Republican Party and Donald Trump's populism, and eventually using it to try to usher in Communism, via a World Economic Forum/Black Rock policy and vision.
2
1
14
1,678
I’ve battled tested both @bot and @Muse. Pending Duet to be available in SG. One potential I haven’t managed to assess is @ManusAI’s Cue. Anyone has an invite? 🙏🏻
3
12
1,056
From early thesis to national infrastructure. When we started building @cludeproject, our core thesis around KYA (Know Your Agent) was simple: as agents take on financial transactions and sensitive tasks, we need a way to verify who they are and what they are mandated to do. We can all agree that a static ID isn't enough, memory is a critical piece of that identity fingerprint. Watching MP Tin Pei Ling and MOS Jasmin Lau discuss a national digital identity registry ("Singpass for AI agents") and interoperable standards in Parliament is massive validation for where this space is heading and what we've been building. Thrilled to be working closely with the relevant bodies on building out Singapore’s Agent Identity framework, bringing verifiable memory into the equation. Let's get to building the standard Video credit to @straits_times
.@cludeproject is collaborating with Singapore's regulatory bodies to define a new memory standard for AI agents, mirroring the impact of @AnthropicAI's MCP By providing a clear audit trail of what an agent did and why, they are bridging the gap between AI innovation and tight financial regulations. $CLUDE
9
6
5
55
8,506
One of the key variables we’ve pioneered at Clude is leveraging the memory layer as a core piece of an agent's verifiable identity. An agent isn't just its base model; it’s their context, state, and history. Similar to how we are shaped to who we are, with our unique memory and context. W!
1
12
518
Pretty similar approach to Clude and what we built 6 months ago. Interesting to understand how they handle memory decay
We decoupled memory from the model. Living Brain is one persistent brain for ChatGPT, Claude, Cursor, Codex and your own agents. Write in one, retrieve in any. 92.2% on LoCoMo. 464ms median retrieval. Up to 22.6x fewer tokens per question. LLMs have parameters. Living Brain has plasticity.
3
1
1
12
951
Honestly, I can't envision my day to day nowadays without @bot
7
15
1,613
GM - we just crossed 10K active users, processing over 30M memory nodes. Momentum
17
6
5
39
4,091
This was when I audited my project and found a ton of security vulnerabilities, many of which are shared with other platforms that use the same libraries and the same frameworks. I've already passed these on to the relevant bodies, so I won't be publishing details. But I'm bringing it up because of what's happening right now. 207 crypto hacks in the first half of this year, up from 83 last year. $2.69B lost by September (excluding what just happened), already more than all of 2023. Bitget lost about $387M, and now Binance is dealing with their the recent "hack" Most of it isn't smart contract bugs. It's the stuff that most overlook: third-party tools, stolen credentials, shared dependencies. So please..take security seriously. It's not something you add later.
Clude is back up What I thought was a simple patch actually revealed many more server side vulnerabilities. AI accelerated our production phase, but it also brings forth a new kind of secops challenge
2
26
1,801
Clude is back up What I thought was a simple patch actually revealed many more server side vulnerabilities. AI accelerated our production phase, but it also brings forth a new kind of secops challenge
4
3
2
26
2,849
Still in the midst of cleaning up in the wake of the Railway and Vercel aftermath. Expect some downtime on Clude. Target to be back up today.
4
2
21
1,552