Samuel Cochran.oci

16.4K posts

Samuel Cochran.oci banner
Samuel Cochran.oci

Samuel Cochran.oci

@sj26

Founding Engineer at @Buildkite

Melbourne, Victoria Katılım Mayıs 2007
774 Takip Edilen1.1K Takipçiler
@melissa
@melissa@melissa·
it seems patently better to buy in to persistence as opposed to the belief that losing results from randomness or unfairness, and never that you were outplayed
@melissa tweet media
English
2
7
213
12.9K
@melissa
@melissa@melissa·
one time as a kid, my dad told me about some old research he'd read he said, maybe asian kids are better at math because they just try 50% longer
@melissa tweet media
English
97
702
15.3K
1.1M
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
Trying to read a markdown document on @GitHub this morning and it keeps refreshing and scrolling back to the top. What are they doing with their frontend???
English
0
0
0
72
Ty
Ty@TheDyla34646123·
@NAP_benkimchi What’s it called on steam?
English
2
0
0
3K
Benkimchi
Benkimchi@NAP_benkimchi·
GAMEPLAY:You can roughly think of it as a version of TAUR with a movable base. ARTSTYLE: 90s Mecha anime #gamedev #indiegame #ゲーム制作 #ゲーム開発
English
57
480
4.4K
283K
Massimo
Massimo@Rainmaker1973·
This is a Wagyu Steak Rice Bowl restaurant in Osaka. The A5 Tadaki brand uses only Wagyu beef, grilled over charcoal, and served with tare sauce. This restaurant is located near Dotonbori.
English
178
482
7.2K
1.6M
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
The crisis of software quality was already coming. Coding agents have just accelerated everything. Folks shipping quality are still shipping quality but faster. Those in crisis are unravelling faster. The problems are systemic. So must be the solutions.
English
0
0
3
92
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
@dhh @buildkite (We still have int primary keys for database and legacy reasons but they’re an implementation detail, everything has a uuid representing its true identity, public apis have only ever exposed uuids, and many models have uuid primary keys only.)
English
0
0
0
77
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
@dhh Same as @Buildkite. I’m hoping we can bridge our central service into self hosted components with distributed key generation etc. Hardest part is sequential build numbers 😅
English
1
0
0
151
DHH
DHH@dhh·
Fun fact: Fizzy is using UUIDv7 primary keys. Plan is to offer import/export between our SaaS hosting and people's own on-prem installations. Start on SaaS, grab your own SQLite DB, then run it yourself (or the other way!). Will upstream to Rails soon. github.com/basecamp/fizzy…
English
19
15
430
47.2K
parks
parks@parkersity_9·
Without naming your job, tell me something you say 20 times per day at work
English
11.6K
387
10.6K
12.1M
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
For a product about exploring time series data, @datadoghq is truly terrible at time. Copy/pasting within the application confuses it.
Samuel Cochran.oci tweet mediaSamuel Cochran.oci tweet media
English
0
0
1
105
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
Service meshes are the middle management of software architecture.
English
0
0
1
54
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
@markjaquith This is also how you trick a toddler out of a temper tantrum, start saying things that are wrong - oh look at that blue apple, what a nice red banana, has you seen this purple orange??
English
0
0
1
54
Mark Jaquith
Mark Jaquith@markjaquith·
Tip for engineers working with talented but time-constrained designers: Instead of asking them to design something and waiting weeks, just slap something together. It will literally jolt them awake like a cartoon and they’ll design something 10× better, same day.
dax@thdxr

it felt bad to me that we all copied the same shimmer loader from each other it's also hard to see for some people to see esp if they're in light mode swapped it out for something simpler but i like it more david might wake up and make me change it

English
25
17
881
84.4K
Jane Manchun Wong
Jane Manchun Wong@wongmjane·
Looks like Google reverse-engineered AirDrop’s protocol and came up with their own implementation! This is so cool! Now, if Apple deliberately breaks the interoperability and limits it to Apple devices only… I think Google might be able to build an antitrust case out of it
Android@Android

Sharing moments shouldn’t depend on the phone you have. Starting today with the Pixel 10 family, Quick Share now works with AirDrop, making secure file transfers between Android phones and iPhones more seamless. This builds on our commitment to cross-OS compatibility to bridge the gap between ecosystems.  We built this feature with security top of mind from day one. Learn more ↓ blog.google/products/andro…

English
139
768
13.3K
1M
Ivan Jukic
Ivan Jukic@ivjukic·
@wongmjane Nobody mentions that EU adopted new specification decisions that required Apple to adopt new interoperable wireless standards, starting in this year’s iOS 26 release. So basically every phone will be able transfer data with iPhone.
English
1
0
2
675
Samuel Cochran.oci
Samuel Cochran.oci@sj26·
Yeah our tools already detect and retry flaky tests automatically etc so if a build fails it requires diagnosis beyond a repeatable pattern. That’s why it’s tricky. What the human would do is dive deeply into the build result and its artefacts to understand the error and how to fix it. This is a difficult problem to solve generally. Pipelines, build graphs, etc, vary wildly. Solving this specifically is better, as a human at a terminal with an agent, or an agent in a coding environment of its own trying to push a PR. Neither of those should live within CI, imho, they are adjacent.
English
0
0
0
23
matt swanson 😈
matt swanson 😈@_swanson·
I think a good angle is: what would a human do and then how could an LLM mirror that? So if CI fails, what does a human do: - I see the alert - I check the error to see if it seems like a flake, if so I re-run - If not, I pattern-match/search my brain to see if this is a known type of failure - I leave a comment on the alert in Slack that says "Flaky; re-running" or "Weird error, taking a look" So instead, could an LLM be inserted so that: if it seems flaky (how to detect? dunno but you could give an LLM examples of common flaky tests or look at what builds re-ran without code changes), just re-run it for me. If it still fails, maybe the Slack alert has a note at the bottom like (We tried re-running this and it still failed)
English
1
0
0
57
matt swanson 😈
matt swanson 😈@_swanson·
I would love to read much more about using LLMs in product features and much less about using LLMs to code the features
English
3
3
27
2.3K