Hacker News: mikeshi42

New comment by mikeshi42 in "Show HN: Make your logo extra bright on HDR screens"

mikeshi42 — Sun, 23 Aug 2026 00:24:30 +0000

I’ve always wondered why it happened, assumed it was a rendering bug.

fwiw it seems like “reduce white point” overrides any of these effects - which is good.

New comment by mikeshi42 in "Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users"

mikeshi42 — Fri, 07 Aug 2026 00:18:38 +0000

High CACs are supported by high LTVs, that is about the right ballpark for things like lawyers or even plumbers.

Building a better MCP server and proving it

mikeshi42 — Thu, 06 Aug 2026 04:37:28 +0000

Article URL: https://clickhouse.com/blog/benchmarking-the-clickstack-mcp-server-with-hdx-evals

Comments URL: https://news.ycombinator.com/item?id=49192503

Points: 1

# Comments: 0

Your team will build the best SRE agent

mikeshi42 — Thu, 30 Jul 2026 16:38:45 +0000

Article URL: https://clickhouse.com/blog/the-future-of-observability-not-one-proprietary-ai-agent-thousands-by-teams

Comments URL: https://news.ycombinator.com/item?id=49112393

Points: 1

# Comments: 0

We built an MCP server for your SRE agent

mikeshi42 — Wed, 29 Jul 2026 14:25:29 +0000

Article URL: https://clickhouse.com/blog/benchmarking-the-clickstack-mcp-server-with-hdx-evals

Comments URL: https://news.ycombinator.com/item?id=49097958

Points: 10

# Comments: 2

ClickCannon: Building a Tool for Benchmarking ClickHouse

mikeshi42 — Tue, 28 Jul 2026 17:22:43 +0000

Article URL: https://clickhouse.com/blog/building-clickcannon-a-tool-for-benchmark-clickhouse

Comments URL: https://news.ycombinator.com/item?id=49087098

Points: 2

# Comments: 0

New comment by mikeshi42 in "GPT-5.6"

mikeshi42 — Fri, 10 Jul 2026 04:39:02 +0000

I've been using both and as far as I can tell with ccusage, the $ equivalent budgets are about the same between them now. This may have been true before anthropic doubled their quotas and openai 2x promo expired last month.

The future of observability won't be one proprietary AI agent

mikeshi42 — Mon, 29 Jun 2026 16:11:48 +0000

Article URL: https://clickhouse.com/blog/the-future-of-observability-not-one-proprietary-ai-agent-thousands-by-teams

Comments URL: https://news.ycombinator.com/item?id=48721122

Points: 1

# Comments: 0

New comment by mikeshi42 in "A successful Japanese trial of a ramjet engine designed for Mach‑5 aircraft"

mikeshi42 — Tue, 26 May 2026 02:36:50 +0000

I have _yet_ to hit a time where TSA can make multiple hours disappear. Precheck w/ touchless ID lines are virtually empty at most airports, the actual security screen itself is quite fast given almost nothing needs to be removed from your bag these days. I still tend to arrive early, but I don't mind getting work done at the airport, especially at a lounge - though I've arrived very close to departure other times and still make it to the gate with plenty to spare.

On international returns, both Global Entry or MPC lines are virtually empty when I arrive (SFO)

The worst part is international arrivals in foreign countries, where immigration can soak up a lot of time, and you have no choice but to stand in line. Luckily I don't have to fly internationally too many times a year.

New comment by mikeshi42 in "Appearing productive in the workplace"

mikeshi42 — Thu, 07 May 2026 16:41:02 +0000

The irony is when I was sanity checking with ChatGPT - it caught the inaccuracies on its own.

New comment by mikeshi42 in "Appearing productive in the workplace"

mikeshi42 — Thu, 07 May 2026 05:39:16 +0000

While I agree with some of these observations - the research cited in the article really do not match the claims at all from what I can tell.

> An NBER study of support agents [2] found generative AI boosted novice productivity by about a third while barely helping experts. Harvard Business School researchers found the same pattern in consulting work [3].

The first work cited was a research study on GPT-3(!) from 2020. Which is a barely coherent model relative to today's SOTA.

The second HBS research study literally finds the opposite of what's claimed:

> we observed performance enhancements in the experimental task for both groups when leveraging GPT-4. Note that the top-half-skill performers also received a significant boost, although not as much as the bottom-half-skill performers.

Where bottom-half skilled participants with AI outperformed top-half skilled participants without AI. (And top-half skilled participants gained another 11% improvement when pared with AI). Again, GPT-4 model intelligence (3 years ago) is a far cry from frontier models today.

New comment by mikeshi42 in "LinkedIn is searching your browser extensions"

mikeshi42 — Thu, 02 Apr 2026 20:43:47 +0000

I’ve seen fake accounts created by bad actors attempting to pose as others for gaining remote employment. It’s possible that is what was happening, and the takedown was from LI taking down the profile from the bad actor.

Other times they would just link to real LinkedIn profiles, but the LinkedIn profile will say that they’re not actively looking and are a victim of id fraud basically.

It’s been a huge issue spotting candidates falsifying information since remote work took off unfortunately. They payout is if they can get at least 1 or 2 paychecks before being found out, they’ve made a good profit.

Claude Opus 4.6 extends LLM pareto frontier

mikeshi42 — Sat, 07 Feb 2026 17:07:51 +0000

Article URL: https://michaelshi.me/pareto/

Comments URL: https://news.ycombinator.com/item?id=46925432

Points: 2

# Comments: 0

New comment by mikeshi42 in "LLM Pareto Frontier"

mikeshi42 — Sun, 18 Jan 2026 20:40:31 +0000

I'm actually curious on the evolution of models as well over time - I created a "created at" group by option to try to broadly visualize how models over time (half calendar years) drift in terms of pareto efficiency. There's a clear trend going up which is good.

I hope to build some better visualizations down the line on model evolution instead of just hacking it onto the current scatterplot.

LLM Pareto Frontier

mikeshi42 — Sun, 18 Jan 2026 19:52:48 +0000

Article URL: https://michaelshi.me/pareto/

Comments URL: https://news.ycombinator.com/item?id=46671499

Points: 4

# Comments: 2

ClickHouse Acquires LibreChat

mikeshi42 — Thu, 06 Nov 2025 22:02:17 +0000

Article URL: https://clickhouse.com/blog/librechat-open-source-agentic-data-stack

Comments URL: https://news.ycombinator.com/item?id=45841017

Points: 1

# Comments: 0

LLMs can't solve production issues

mikeshi42 — Wed, 17 Sep 2025 18:35:59 +0000

Article URL: https://clickhouse.com/blog/llm-observability-challenge

Comments URL: https://news.ycombinator.com/item?id=45279656

Points: 3

# Comments: 1

Can LLMs replace on-call SREs?

mikeshi42 — Thu, 14 Aug 2025 17:12:57 +0000

Article URL: https://clickhouse.com/blog/llm-observability-challenge

Comments URL: https://news.ycombinator.com/item?id=44903013

Points: 1

# Comments: 1

New comment by mikeshi42 in "The current state of LLM-driven development"

mikeshi42 — Sun, 10 Aug 2025 16:34:19 +0000

Yes, though that just means the probability of success is a function of not only user input but also the model version.

Slot machines on the other hand are truly random and success is luck based with no priors (the legal ones in the US anyways)

New comment by mikeshi42 in "The current state of LLM-driven development"

mikeshi42 — Sun, 10 Aug 2025 11:36:31 +0000

There's plenty of evidence that good prompts (prompt engineering, tuning) can result in better outputs.

Improving LLM output through better inputs is neither an illusion, nor as easy as learning how to google (entire companies are being built around improving llm outputs and measuring that improvement)