Joe Gibbons

3.5K posts

Joe Gibbons

Joe Gibbons

@JZGibbons

Katılım Mart 2021
2.3K Takip Edilen278 Takipçiler
Everlier
Everlier@Everlier·
@atmoio This is how you spot leaders that didn't use the tech deeply enough to understand its limits. Whelp.
English
2
1
36
3.3K
Mo
Mo@atmoio·
tldr: ClickUp is hosting a company-wide Hunger Games where if you can figure out how the hell to make AI work you’ll win a million dollars.
Zeb Evans@DJ_CURFEW

Today we reduced headcount by 22%. The business is the strongest it's ever been. So I think it's important to be direct about what I'm seeing and why. First, I made this decision and I own it. I did it because the way to operate at the highest level of productivity is changing, and to win the future, ClickUp needs to change with it. Second, this wasn't about cutting costs. Most savings from this change will flow directly back into the people who stay. We'll be introducing million-dollar salary bands. If you create outsized impact using AI, you'll be paid outside of traditional bands. Most importantly, I have the deepest gratitude for those affected. We're doing this from a position of strength specifically so we can take care of people properly. Everyone affected receives a package aimed at honoring their contributions and easing the transition. I only see two options: wait for this to play out gradually in the market or be honest about what I'm seeing and act proactively. THE 100X ORGANIZATION The primary change is that we're restructuring around what I call 100x org. The goal is 100x output. The roles required to build at the highest level are fundamentally different than they were a year ago. Incremental improvements to existing systems won't get us there. We need new ones. That means creating enough disruption to rebuild rather than iterate on what's already broken. The common narrative is that AI makes everyone more productive. It doesn't. Many of the workflows of today, if left unchanged, create bottlenecks in AI systems. These roles will evolve. But waiting for that to happen naturally means falling behind now. The 100x org is actually heavily dependent on people - infinitely more than today. This is only possible with 10x people that have embraced and adopted new ways of working. THE BUILDERS, AGENT MANAGERS, AND FRONT-LINERS — THE BUILDERS: 10X ENGINEERS I don't think most companies have internalized what's actually happening with AI in engineering. The common narrative is that AI makes all engineers more productive. That may be true in isolation, but at an organization level - that is the farthest thing from reality. Here's what we've validated recently at ClickUp: the great engineers, the ones who can orchestrate, architect, and review, are becoming 100x engineers. They're not writing code. They're directing agents that write code. The skill is judgment. AI makes the best engineers wildly more productive, and everyone else using AI slows these engineers down. Think about it - the bottlenecks are (1) orchestration - telling AI what to do, and (2) reviewing - what AI did. Everything is leapfrogged and no longer needed. So who do you want orchestrating and reviewing code? And how do you want your best engineers to spend their time? If your best engineers are spending time reviewing other people's code, then this is inherently an inefficient bottleneck. These engineers can review their agent's code much faster than reviewing human code. The new world is about enabling your 10x engineers to become 100x. The wrong strategy is to push every engineer to use infinite tokens. Companies doing this are celebrating 500% more pull requests. But customer outcomes don't match the volume of code being generated. I call this the great reckoning of AI coding, and every company will face this soon if not already. More code is just another bottleneck to the best engineers, and ultimately to your company's impact as well. — THE BUILDERS: 10X PRODUCT MANAGERS Product management and design roles are merging. Designers that have customer focus, become more like product managers. And product managers that have intuition for UX become more like designers. The bottleneck of user research is gone. It takes us just one mention of an agent to kickoff research and analyze results. The bottleneck of product <> design iteration is also gone. The product builder iterates on their own, along with agents and skills that ensure alignment with quality and strategy. Also controversial today - I believe that the wrong strategy is to have your PMs shipping code - that just introduces another bottleneck that the best engineers will waste their time on. To be clear, PMs should be coding but they should do this in a playground to iterate, validate, and scope. That code should not go to production. Everything outside of managing systems, orchestrating AI, and reviewing output becomes a bottleneck. That's why the other roles that are critical along with these are the systems managers (to reduce bottlenecks) along with a bottleneck you can't replace - customer meeting time. — THE SYSTEM MANAGERS Ironically, the people that automate their jobs with AI will always have a job. They become owners of the AI systems - agent managers. We have many examples of these people at ClickUp. The underlying systems in which we operate are absolutely critical to get right. I think most companies are delusional to think they can iterate on existing systems and compete in this new world. You must create enough disruption so that old systems are deprecated entirely. If there's any definition for 'AI native' that's what it is. — THE FRONT-LINERS In a world that will become saturated with AI communication, the human touch will matter more than anything to customers. This is a bottleneck that you shouldn't replace - even when agents are high enough quality to do video meetings. One-on-one meeting time with customers is something that shouldn't be automated. The systems around the meetings should be - so that front-liners spend nearly 100% of their time with customers. REWARDING 100X IMPACT In a world where companies are able to do so much more with less, where does that excess money go? In our case, much of the savings in this new operating model will flow directly back to those that enabled it. We must reward people that create productivity accordingly. This aligns incentives on both sides. Plus, in a world where your best people create 100x impact, you can't afford to lose them. You should aim to retain these employees for decades. The context they have and their ability to efficiently orchestrate and review will be nearly impossible to replace. Compensation bands of today should be thrown out the door. We're introducing $1 million cash/year salary bands with a path available to nearly everyone in the company if they produce 100x impact by creating or managing AI systems. THE FUTURE Nearly every company will make changes like these. The ones that do it proactively will define what comes next. The future is not fewer people. It's different work, new roles, and better rewards for those who embrace it. We're already seeing entirely new roles emerge, like Agent Managers, that didn't exist a year ago. ClickUp is positioning to lead this shift, not just internally, but for our customers too. I've never been more certain about where we're headed.

English
110
166
3.2K
258.1K
clem 🤗
clem 🤗@ClementDelangue·
the reason is marketing!
Patrick OShaughnessy@patrick_oshag

Interesting that in three of my recent conversations with Krishna at Anthropic, @dylan522p, and @GavinSBaker, each said that frontier tokens are capturing the majority of the economic value. Gavin: "An overwhelming amount of the economic returns to AI at the model layer have been at the frontier. That's surprising to me, and I think it's been surprising to a lot of people. This is one of the most important questions to be answered, and you need to have a hypothesis on it as an investor." Krishna: "We think the returns to frontier intelligence are extremely high. Customers invest really heavily in more tokens with the newer models. The ones at the frontier clearly are capturing this economic value, driving meaningful ROI for customers. The returns to frontier intelligence are not slowing down." Dylan: "No one gives a crap about GPT-4 class models. They want the frontier because the frontier lets them create the economically valuable things."

English
7
2
26
7.5K
Joe Gibbons retweetledi
wd 🔺
wd 🔺@populartourist·
llama.cpp release b9235 added some new toys for boosting inference. Benchmarked Qwen3.6 27B on an RTX 5090 with llama.cpp, using speculative n-gram tuning across 10k generated tokens tests. Increasing --spec-ngram-map-k4v-size-m scaled decode throughput (predicted_per_second) up to ~7x faster accepted output token generation. A follow-up 7x50k token generation tests on k4v64 and k4v96 samples confirmed the sustained 10k-token performance, making k4v96 winner. k4v128 was tested too, but less stable against k4v96 in the 7x50k token run, so it was removed from the charts. Real-world results remain anecdotal, albeit k4v96 showed a much lower acceptance rate than traditional --spec-draft-n-max 3 while still producing faster evaluation speeds - so the trade-off seems to be worth it. Flags in comments below for the k4v96 tested sample.
wd 🔺 tweet media
English
19
30
314
45.6K
Joe Gibbons
Joe Gibbons@JZGibbons·
@antigravity That's fantastic but is it trained to be far-left over reality?
English
0
0
0
416
Google Antigravity
Google Antigravity@antigravity·
Introducing Gemini 3.5 Flash ⚡️ Normally, Gemini 3.5 Flash is 4x faster than other models with frontier performance. For a limited time, Antigravity is serving it 12x faster thanks to custom inference tricks, delivering incredible speed for your workflows. 🚀 See the performance in this demo: generating pixel art from photos, orchestrating multi-agent workflows to write and register sprites, and spawning browser subagents to auto-test rendering: 👇
English
97
85
1.2K
129.4K
Ejaaz
Ejaaz@cryptopunk7213·
i was completely wrong about cursor composer 2.5 is an amazing model and i’m convinced the model harness is as important as training. this bodes well for the cursor + spacexai combo they’re the only one with a shot at joining the leaders in frontier coding. wrote some thoughts
Ejaaz@cryptopunk7213

x.com/i/article/2056…

English
9
8
296
20.9K
Joe Gibbons
Joe Gibbons@JZGibbons·
@jun_song closing in on a sweet spot of cloud pricing vs locallllm, I'm going for the latter - forgetting the East v West, capitalist v communist politics - the commies are giving me quality models for free! I'll take em, with abliteration!
English
0
0
0
17
Jun Song
Jun Song@jun_song·
Every AI frontier labs are cutting our usage limit without notice. Because it’s legal, and inference cost is skyrocketing. It’s not a surprise. Get Local LLMs now.
J J@jturntdev

OpenAI have secretly adjusted our limits. Last week before limit reset. I was using Xhigh all day. 5 day straight i couldn’t get my usage below 55% weekly usage. Since Yesterday, I’ve done 40% of my quota, out of nowhere. So whats going on ? @thsottiaux @sama @OpenAIDevs

English
9
6
66
4.5K
Joe Gibbons
Joe Gibbons@JZGibbons·
@jun_song @Maciej25956571 very much agree with this, get in now on a 3090 or more if you can afford it. It may be slower and need a lot of prodding but you can keep your job for a long time if you agentic code it/DSPy/script it right, lots of thinking GBNF/templates out there to get thing working better
English
0
0
0
23
Jun Song
Jun Song@jun_song·
@Maciej25956571 That’s why we should buy before everything skyrockets. I’m expecting current $200 subs to be $1000 (5x) by the end of year.
English
2
0
2
119
Jun Song
Jun Song@jun_song·
What makes you think Cloud AI will get cheaper? (I still get lots of comments from non-followers) > Grok is training 1.5T model (currently 0.5T) > Mythos has x3 parameters than Opus > GPT parameters getting larger It’s simple math, they need to spend x3 on inference to serve.🧵
English
10
0
59
5.8K
Joe Gibbons
Joe Gibbons@JZGibbons·
@LLMJunky X should include List members here, I may not follow you but you're in my "AI" list - just to keep things segmented on my interests
English
1
0
1
14
am.will
am.will@LLMJunky·
Show me your active follower counts.
am.will tweet media
English
10
0
15
1.2K
Joe Gibbons retweetledi
Nick Timothy MP
Nick Timothy MP@NJ_Timothy·
They think they can do what they like without a mandate. The problem isn't Starmer. It's Labour.
Nick Timothy MP tweet media
English
480
2.7K
11K
99.3K
Joe Gibbons retweetledi
The Rabbit Hole
The Rabbit Hole@TheRabbitHole·
The Woke Mind Virus in Academia
The Rabbit Hole tweet media
English
276
1K
4.9K
9.3M
Joe Gibbons retweetledi
Joe Gibbons
Joe Gibbons@JZGibbons·
@stevibe Cheers. This is great but a little issue when I have all the Bench Packs installed, the New Tan dialog where you pick one out doesn't show the Open Benchmark button - might simply be my font sizes and there is a workaround to search for the Bench Pack. Great though, thanks.
English
0
0
0
27
stevibe
stevibe@stevibe·
BenchLocal v0.2.5 is out! > The big one: repeated test runs with majority voting (1, 3, 5, 7, or 9 runs per test). > Plus error classification, retry actions, per-scenario timings & more. github.com/stevibe/BenchL…
stevibe tweet media
English
5
3
36
2.7K
Joe Gibbons retweetledi
Emma P
Emma P@ThisIsEmmaPx·
Migrants from Former RAF Wethersfield spotted Cash Trapping today in Chelmsford. We keep raising awareness locally even if I am supposed to switching off in ibiza.
Emma P tweet media
English
205
3.2K
8.3K
587.7K
Joe Gibbons
Joe Gibbons@JZGibbons·
@levelsio need a localllm version of this with genuine logs of prompts, 12, 16, 24GB models - it would be fun to see what can be done with a clever harness (prizes aq lot lower in $ though)
English
0
0
0
94
@levelsio
@levelsio@levelsio·
🏆 The Vibe Jam of 2026 sponsored by @cursor_ai + @boltdotnew + @heyglif + @tripoai is now being judged! Me and @s13k_ are now judging the ~1000 games of When either one of us likes a game, it goes on to the next round, so it's like a first stage @s13k_ built an awesome judging system that lets us just play the games in the browser (most work in an iframe) and then immediately score them Then in the second round, we send the remaining games to the other judges: - @timsoret from @oddtalesgames - @ericzakariasson from @cursor_ai - @NicolaManzini from VibeSail - Alessandro from @OverJumpRally The prizes are: - First place: $25,000 - Second place: $10,000 - Third place: $5,000 19 games done, 926 to go!
@levelsio tweet media
@levelsio@levelsio

And.....✅ DONE! The Vibe Jam of 2026 sponsored by @cursor_ai + @boltdotnew + @heyglif + @tripoai is officially closed 🕹️ 945 games submitted 🛝 242,212 players 👁️ ~12 million views on X Now me and @s13k_ will start the judging process, probably pre-vetting games first with some help from AI (like if the games load at all) and then they go on to all the judges I want to thank everyone who participated! ❤️ There's only 3 cash prizes but even if you don't win, I hope you all had fun creating things, which is the best part of AI for me, it lets me create things I could never have dreamt of making before It's already clear to me from the submissions that AI's ability to help you create beautiful and fun games has progressed a lot, last year's games looked clunky and basic, this year's games are starting to look like stuff you could find on Steam There's no specific deadline for when judging is done but we'll try to be as fast as possible, last year it took 2 weeks I think! THANK YOU!!!

English
22
10
209
41.6K
Joe Gibbons
Joe Gibbons@JZGibbons·
@DealsForge @cjzafir I found that too, they're so fast that it's so easy to correct any mistakes. The ELM concept does sound intriguing for a specialized coding model, I hope we get OSS versions of those.
English
0
0
0
23
Ucav
Ucav@DealsForge·
@cjzafir Smaller models make feedback loop honest. You see faster if the dataset, evals, prompting, task definition or deployment path are actually solid. But scaling up should amplify a working loop, not hide a broken one !!
English
1
0
3
297
CJ Zafir
CJ Zafir@cjzafir·
If you love fine-tuning open-source models (like me), then listen. > Start with 1B, 2B, 4B, and 8B models. (Don't start with a 27B model or bigger at first.) > Use WebGPU providers. I use Google Colab Pro for any model smaller than 9B. A single A100 80GB costs around $0.60/hr, which is cheap. Enough for small models. > Don’t buy GPUs unless you fine-tune 7 to 10 models. You'll understand the nitty-gritty in the process. > Use Codex 5.5 × DeepSeek v4 Pro to create datasets. Codex to plan, DeepSeek v4 Pro to generate rows. > Use Unsloth's instruct models as a base from Hugging Face. Yes, there are others too, but Unsloth also provides fast fine-tuning notebooks. > Use Unsloth's fine-tuning notebooks as a reference. Paste them into Codex, and Codex will write a custom notebook with the configs you need. > Spend 1 day learning about: - SFT (supervised fine-tuning) - RL training (GRPO, DPO, PPO, etc.) - LoRA / QLoRA training - Quantization and types - Local inference engines (llama.cpp) - KV cache and prompt cache > Just get started. Claude, Codex, and ChatGPT can design a step-by-step plan for how you can fine-tune your first AI model. Future tech is moving toward small 5B to 15B ELMs (Expert Language Models) rather than general 1T LLMs. So fine-tuning is an important skill that anyone can acquire today. Tune models, test them, use them. Then fine-tune for companies and make a career out of it. (Companies pay $50k+ to fine-tune models on their data so they can get personalized AI models.) Shoot your questions below. I'll be sharing in-depth raw findings about this topic in the coming days.
CJ Zafir tweet media
English
97
317
2.5K
185.4K
Joe Gibbons
Joe Gibbons@JZGibbons·
@cjzafir I used unsloth as a test to train the 0.8B Qwen3.5, obviously it's rapid but also quite useful. I used GLM-5 to generate a 4000 training inputs and 250 validation cases from my codebases at work, it generates code based on those even on a CPU at an amazing rate.
English
0
0
0
267